diff options
author | marha <marha@users.sourceforge.net> | 2011-01-28 08:01:47 +0000 |
---|---|---|
committer | marha <marha@users.sourceforge.net> | 2011-01-28 08:01:47 +0000 |
commit | dc0f390b45a3037c85163b8a9e945a31f5756690 (patch) | |
tree | 317b992cf88d5a7df8c9b9c7c419aef42246fa3f /libxcb/src/c_client.py | |
parent | 987904f9692c374b81547ebe0bd3bad5a279db48 (diff) | |
download | vcxsrv-dc0f390b45a3037c85163b8a9e945a31f5756690.tar.gz vcxsrv-dc0f390b45a3037c85163b8a9e945a31f5756690.tar.bz2 vcxsrv-dc0f390b45a3037c85163b8a9e945a31f5756690.zip |
libfontenc libXau libxcb mesalib git update 28 jan 2011
Diffstat (limited to 'libxcb/src/c_client.py')
-rw-r--r-- | libxcb/src/c_client.py | 4625 |
1 files changed, 2315 insertions, 2310 deletions
diff --git a/libxcb/src/c_client.py b/libxcb/src/c_client.py index fa65c1baa..a10b3f1d8 100644 --- a/libxcb/src/c_client.py +++ b/libxcb/src/c_client.py @@ -1,2310 +1,2315 @@ -#!/usr/bin/env python
-from xml.etree.cElementTree import *
-from os.path import basename
-import getopt
-import sys
-import re
-
-# Jump to the bottom of this file for the main routine
-
-# Some hacks to make the API more readable, and to keep backwards compability
-_cname_re = re.compile('([A-Z0-9][a-z]+|[A-Z0-9]+(?![a-z])|[a-z]+)')
-_cname_special_cases = {'DECnet':'decnet'}
-
-_extension_special_cases = ['XPrint', 'XCMisc', 'BigRequests']
-
-_cplusplus_annoyances = {'class' : '_class',
- 'new' : '_new',
- 'delete': '_delete'}
-_c_keywords = {'default' : '_default'}
-
-_hlines = []
-_hlevel = 0
-_clines = []
-_clevel = 0
-_ns = None
-
-# global variable to keep track of serializers and
-# switch data types due to weird dependencies
-finished_serializers = []
-finished_sizeof = []
-finished_switch = []
-
-def _h(fmt, *args):
- '''
- Writes the given line to the header file.
- '''
- _hlines[_hlevel].append(fmt % args)
-
-def _c(fmt, *args):
- '''
- Writes the given line to the source file.
- '''
- _clines[_clevel].append(fmt % args)
-
-def _hc(fmt, *args):
- '''
- Writes the given line to both the header and source files.
- '''
- _h(fmt, *args)
- _c(fmt, *args)
-
-# XXX See if this level thing is really necessary.
-def _h_setlevel(idx):
- '''
- Changes the array that header lines are written to.
- Supports writing different sections of the header file.
- '''
- global _hlevel
- while len(_hlines) <= idx:
- _hlines.append([])
- _hlevel = idx
-
-def _c_setlevel(idx):
- '''
- Changes the array that source lines are written to.
- Supports writing to different sections of the source file.
- '''
- global _clevel
- while len(_clines) <= idx:
- _clines.append([])
- _clevel = idx
-
-def _n_item(str):
- '''
- Does C-name conversion on a single string fragment.
- Uses a regexp with some hard-coded special cases.
- '''
- if str in _cname_special_cases:
- return _cname_special_cases[str]
- else:
- split = _cname_re.finditer(str)
- name_parts = [match.group(0) for match in split]
- return '_'.join(name_parts)
-
-def _cpp(str):
- '''
- Checks for certain C++ reserved words and fixes them.
- '''
- if str in _cplusplus_annoyances:
- return _cplusplus_annoyances[str]
- elif str in _c_keywords:
- return _c_keywords[str]
- else:
- return str
-
-def _ext(str):
- '''
- Does C-name conversion on an extension name.
- Has some additional special cases on top of _n_item.
- '''
- if str in _extension_special_cases:
- return _n_item(str).lower()
- else:
- return str.lower()
-
-def _n(list):
- '''
- Does C-name conversion on a tuple of strings.
- Different behavior depending on length of tuple, extension/not extension, etc.
- Basically C-name converts the individual pieces, then joins with underscores.
- '''
- if len(list) == 1:
- parts = list
- elif len(list) == 2:
- parts = [list[0], _n_item(list[1])]
- elif _ns.is_ext:
- parts = [list[0], _ext(list[1])] + [_n_item(i) for i in list[2:]]
- else:
- parts = [list[0]] + [_n_item(i) for i in list[1:]]
- return '_'.join(parts).lower()
-
-def _t(list):
- '''
- Does C-name conversion on a tuple of strings representing a type.
- Same as _n but adds a "_t" on the end.
- '''
- if len(list) == 1:
- parts = list
- elif len(list) == 2:
- parts = [list[0], _n_item(list[1]), 't']
- elif _ns.is_ext:
- parts = [list[0], _ext(list[1])] + [_n_item(i) for i in list[2:]] + ['t']
- else:
- parts = [list[0]] + [_n_item(i) for i in list[1:]] + ['t']
- return '_'.join(parts).lower()
-
-
-def c_open(self):
- '''
- Exported function that handles module open.
- Opens the files and writes out the auto-generated comment, header file includes, etc.
- '''
- global _ns
- _ns = self.namespace
- _ns.c_ext_global_name = _n(_ns.prefix + ('id',))
-
- # Build the type-name collision avoidance table used by c_enum
- build_collision_table()
-
- _h_setlevel(0)
- _c_setlevel(0)
-
- _hc('/*')
- _hc(' * This file generated automatically from %s by c_client.py.', _ns.file)
- _hc(' * Edit at your peril.')
- _hc(' */')
- _hc('')
-
- _h('/**')
- _h(' * @defgroup XCB_%s_API XCB %s API', _ns.ext_name, _ns.ext_name)
- _h(' * @brief %s XCB Protocol Implementation.', _ns.ext_name)
- _h(' * @{')
- _h(' **/')
- _h('')
- _h('#ifndef __%s_H', _ns.header.upper())
- _h('#define __%s_H', _ns.header.upper())
- _h('')
- _h('#include "xcb.h"')
-
- _c('#include <stdlib.h>')
- _c('#include <string.h>')
- _c('#include <assert.h>')
- _c('#include "xcbext.h"')
- _c('#include "%s.h"', _ns.header)
-
- if _ns.is_ext:
- for (n, h) in self.imports:
- _hc('#include "%s.h"', h)
-
- _h('')
- _h('#ifdef __cplusplus')
- _h('extern "C" {')
- _h('#endif')
-
- if _ns.is_ext:
- _h('')
- _h('#define XCB_%s_MAJOR_VERSION %s', _ns.ext_name.upper(), _ns.major_version)
- _h('#define XCB_%s_MINOR_VERSION %s', _ns.ext_name.upper(), _ns.minor_version)
- _h(' ') #XXX
- _h('extern xcb_extension_t %s;', _ns.c_ext_global_name)
-
- _c('')
- _c('xcb_extension_t %s = { "%s", 0 };', _ns.c_ext_global_name, _ns.ext_xname)
-
-def c_close(self):
- '''
- Exported function that handles module close.
- Writes out all the stored content lines, then closes the files.
- '''
- _h_setlevel(2)
- _c_setlevel(2)
- _hc('')
-
- _h('')
- _h('#ifdef __cplusplus')
- _h('}')
- _h('#endif')
-
- _h('')
- _h('#endif')
- _h('')
- _h('/**')
- _h(' * @}')
- _h(' */')
-
- # Write header file
- hfile = open('%s.h' % _ns.header, 'w')
- for list in _hlines:
- for line in list:
- hfile.write(line)
- hfile.write('\n')
- hfile.close()
-
- # Write source file
- cfile = open('%s.c' % _ns.header, 'w')
- for list in _clines:
- for line in list:
- cfile.write(line)
- cfile.write('\n')
- cfile.close()
-
-def build_collision_table():
- global namecount
- namecount = {}
-
- for v in module.types.values():
- name = _t(v[0])
- namecount[name] = (namecount.get(name) or 0) + 1
-
-def c_enum(self, name):
- '''
- Exported function that handles enum declarations.
- '''
-
- tname = _t(name)
- if namecount[tname] > 1:
- tname = _t(name + ('enum',))
-
- _h_setlevel(0)
- _h('')
- _h('typedef enum %s {', tname)
-
- count = len(self.values)
-
- for (enam, eval) in self.values:
- count = count - 1
- equals = ' = ' if eval != '' else ''
- comma = ',' if count > 0 else ''
- _h(' %s%s%s%s', _n(name + (enam,)).upper(), equals, eval, comma)
-
- _h('} %s;', tname)
-
-def _c_type_setup(self, name, postfix):
- '''
- Sets up all the C-related state by adding additional data fields to
- all Field and Type objects. Here is where we figure out most of our
- variable and function names.
-
- Recurses into child fields and list member types.
- '''
- # Do all the various names in advance
- self.c_type = _t(name + postfix)
- self.c_wiretype = 'char' if self.c_type == 'void' else self.c_type
-
- self.c_iterator_type = _t(name + ('iterator',))
- self.c_next_name = _n(name + ('next',))
- self.c_end_name = _n(name + ('end',))
-
- self.c_request_name = _n(name)
- self.c_checked_name = _n(name + ('checked',))
- self.c_unchecked_name = _n(name + ('unchecked',))
- self.c_reply_name = _n(name + ('reply',))
- self.c_reply_type = _t(name + ('reply',))
- self.c_cookie_type = _t(name + ('cookie',))
-
- self.need_aux = False
- self.need_serialize = False
- self.need_sizeof = False
-
- self.c_aux_name = _n(name + ('aux',))
- self.c_aux_checked_name = _n(name + ('aux', 'checked'))
- self.c_aux_unchecked_name = _n(name + ('aux', 'unchecked'))
- self.c_serialize_name = _n(name + ('serialize',))
- self.c_unserialize_name = _n(name + ('unserialize',))
- self.c_unpack_name = _n(name + ('unpack',))
- self.c_sizeof_name = _n(name + ('sizeof',))
-
- # special case: structs where variable size fields are followed by fixed size fields
- self.var_followed_by_fixed_fields = False
-
- if self.is_switch:
- self.need_serialize = True
- self.c_container = 'struct'
- for bitcase in self.bitcases:
- bitcase.c_field_name = _cpp(bitcase.field_name)
- bitcase_name = bitcase.field_type if bitcase.type.has_name else name
- _c_type_setup(bitcase.type, bitcase_name, ())
-
- elif self.is_container:
-
- self.c_container = 'union' if self.is_union else 'struct'
- prev_varsized_field = None
- prev_varsized_offset = 0
- first_field_after_varsized = None
-
- for field in self.fields:
- _c_type_setup(field.type, field.field_type, ())
- if field.type.is_list:
- _c_type_setup(field.type.member, field.field_type, ())
- if (field.type.nmemb is None):
- self.need_sizeof = True
-
- field.c_field_type = _t(field.field_type)
- field.c_field_const_type = ('' if field.type.nmemb == 1 else 'const ') + field.c_field_type
- field.c_field_name = _cpp(field.field_name)
- field.c_subscript = '[%d]' % field.type.nmemb if (field.type.nmemb > 1) else ''
- field.c_pointer = ' ' if field.type.nmemb == 1 else '*'
-
- # correct the c_pointer field for variable size non-list types
- if not field.type.fixed_size() and field.c_pointer == ' ':
- field.c_pointer = '*'
- if field.type.is_list and not field.type.member.fixed_size():
- field.c_pointer = '*'
-
- if field.type.is_switch:
- field.c_pointer = '*'
- field.c_field_const_type = 'const ' + field.c_field_type
- self.need_aux = True
- elif not field.type.fixed_size() and not field.type.is_bitcase:
- self.need_sizeof = True
-
- field.c_iterator_type = _t(field.field_type + ('iterator',)) # xcb_fieldtype_iterator_t
- field.c_iterator_name = _n(name + (field.field_name, 'iterator')) # xcb_container_field_iterator
- field.c_accessor_name = _n(name + (field.field_name,)) # xcb_container_field
- field.c_length_name = _n(name + (field.field_name, 'length')) # xcb_container_field_length
- field.c_end_name = _n(name + (field.field_name, 'end')) # xcb_container_field_end
-
- field.prev_varsized_field = prev_varsized_field
- field.prev_varsized_offset = prev_varsized_offset
-
- if prev_varsized_offset == 0:
- first_field_after_varsized = field
- field.first_field_after_varsized = first_field_after_varsized
-
- if field.type.fixed_size():
- prev_varsized_offset += field.type.size
- # special case: intermixed fixed and variable size fields
- if prev_varsized_field is not None and not field.type.is_pad and field.wire:
- if not self.is_union:
- self.need_serialize = True
- self.var_followed_by_fixed_fields = True
- else:
- self.last_varsized_field = field
- prev_varsized_field = field
- prev_varsized_offset = 0
-
- if self.var_followed_by_fixed_fields:
- if field.type.fixed_size():
- field.prev_varsized_field = None
-
- if self.need_serialize:
- # when _unserialize() is wanted, create _sizeof() as well for consistency reasons
- self.need_sizeof = True
-
- # as switch does never appear at toplevel,
- # continue here with type construction
- if self.is_switch:
- if self.c_type not in finished_switch:
- finished_switch.append(self.c_type)
- # special: switch C structs get pointer fields for variable-sized members
- _c_complex(self)
- for bitcase in self.bitcases:
- bitcase_name = bitcase.type.name if bitcase.type.has_name else name
- _c_accessors(bitcase.type, bitcase_name, bitcase_name)
- # no list with switch as element, so no call to
- # _c_iterator(field.type, field_name) necessary
-
- if not self.is_bitcase:
- if self.need_serialize:
- if self.c_serialize_name not in finished_serializers:
- finished_serializers.append(self.c_serialize_name)
- _c_serialize('serialize', self)
-
- # _unpack() and _unserialize() are only needed for special cases:
- # switch -> unpack
- # special cases -> unserialize
- if self.is_switch or self.var_followed_by_fixed_fields:
- _c_serialize('unserialize', self)
-
- if self.need_sizeof:
- if self.c_sizeof_name not in finished_sizeof:
- if not module.namespace.is_ext or self.name[:2] == module.namespace.prefix:
- finished_sizeof.append(self.c_sizeof_name)
- _c_serialize('sizeof', self)
-# _c_type_setup()
-
-def _c_helper_absolute_name(prefix, field=None):
- """
- turn prefix, which is a list of tuples (name, separator, Type obj) into a string
- representing a valid name in C (based on the context)
- if field is not None, append the field name as well
- """
- prefix_str = ''
- for name, sep, obj in prefix:
- prefix_str += name
- if '' == sep:
- sep = '->'
- if ((obj.is_bitcase and obj.has_name) or # named bitcase
- (obj.is_switch and len(obj.parents)>1)):
- sep = '.'
- prefix_str += sep
- if field is not None:
- prefix_str += _cpp(field.field_name)
- return prefix_str
-# _c_absolute_name
-
-def _c_helper_field_mapping(complex_type, prefix, flat=False):
- """
- generate absolute names, based on prefix, for all fields starting from complex_type
- if flat == True, nested complex types are not taken into account
- """
- all_fields = {}
- if complex_type.is_switch:
- for b in complex_type.bitcases:
- if b.type.has_name:
- switch_name, switch_sep, switch_type = prefix[-1]
- bitcase_prefix = prefix + [(b.type.name[-1], '.', b.type)]
- else:
- bitcase_prefix = prefix
-
- if (True==flat and not b.type.has_name) or False==flat:
- all_fields.update(_c_helper_field_mapping(b.type, bitcase_prefix, flat))
- else:
- for f in complex_type.fields:
- fname = _c_helper_absolute_name(prefix, f)
- if all_fields.has_key(f.field_name):
- raise Exception("field name %s has been registered before" % f.field_name)
-
- all_fields[f.field_name] = (fname, f)
- if f.type.is_container and flat==False:
- if f.type.is_bitcase and not f.type.has_name:
- new_prefix = prefix
- elif f.type.is_switch and len(f.type.parents)>1:
- # nested switch gets another separator
- new_prefix = prefix+[(f.c_field_name, '.', f.type)]
- else:
- new_prefix = prefix+[(f.c_field_name, '->', f.type)]
- all_fields.update(_c_helper_field_mapping(f.type, new_prefix, flat))
-
- return all_fields
-# _c_field_mapping()
-
-def _c_helper_resolve_field_names (prefix):
- """
- get field names for all objects in the prefix array
- """
- all_fields = {}
- tmp_prefix = []
- # look for fields in the remaining containers
- for idx, p in enumerate(prefix):
- name, sep, obj = p
- if ''==sep:
- # sep can be preset in prefix, if not, make a sensible guess
- sep = '.' if (obj.is_switch or obj.is_bitcase) else '->'
- # exception: 'toplevel' object (switch as well!) always have sep '->'
- sep = '->' if idx<1 else sep
- if not obj.is_bitcase or (obj.is_bitcase and obj.has_name):
- tmp_prefix.append((name, sep, obj))
- all_fields.update(_c_helper_field_mapping(obj, tmp_prefix, flat=True))
-
- return all_fields
-# _c_helper_resolve_field_names
-
-def get_expr_fields(self):
- """
- get the Fields referenced by switch or list expression
- """
- def get_expr_field_names(expr):
- if expr.op is None:
- if expr.lenfield_name is not None:
- return [expr.lenfield_name]
- else:
- # constant value expr
- return []
- else:
- if expr.op == '~':
- return get_expr_field_names(expr.rhs)
- elif expr.op == 'popcount':
- return get_expr_field_names(expr.rhs)
- elif expr.op == 'sumof':
- # sumof expr references another list,
- # we need that list's length field here
- field = None
- for f in expr.lenfield_parent.fields:
- if f.field_name == expr.lenfield_name:
- field = f
- break
- if field is None:
- raise Exception("list field '%s' referenced by sumof not found" % expr.lenfield_name)
- # referenced list + its length field
- return [expr.lenfield_name] + get_expr_field_names(field.type.expr)
- elif expr.op == 'enumref':
- return []
- else:
- return get_expr_field_names(expr.lhs) + get_expr_field_names(expr.rhs)
- # get_expr_field_names()
-
- # resolve the field names with the parent structure(s)
- unresolved_fields_names = get_expr_field_names(self.expr)
-
- # construct prefix from self
- prefix = [('', '', p) for p in self.parents]
- if self.is_container:
- prefix.append(('', '', self))
-
- all_fields = _c_helper_resolve_field_names (prefix)
- resolved_fields_names = filter(lambda x: x in all_fields.keys(), unresolved_fields_names)
- if len(unresolved_fields_names) != len(resolved_fields_names):
- raise Exception("could not resolve all fields for %s" % self.name)
-
- resolved_fields = [all_fields[n][1] for n in resolved_fields_names]
- return resolved_fields
-# get_expr_fields()
-
-def resolve_expr_fields(complex_obj):
- """
- find expr fields appearing in complex_obj and descendents that cannot be resolved within complex_obj
- these are normally fields that need to be given as function parameters
- """
- all_fields = []
- expr_fields = []
- unresolved = []
-
- for field in complex_obj.fields:
- all_fields.append(field)
- if field.type.is_switch or field.type.is_list:
- expr_fields += get_expr_fields(field.type)
- if field.type.is_container:
- expr_fields += resolve_expr_fields(field.type)
-
- # try to resolve expr fields
- for e in expr_fields:
- if e not in all_fields and e not in unresolved:
- unresolved.append(e)
- return unresolved
-# resolve_expr_fields()
-
-def get_serialize_params(context, self, buffer_var='_buffer', aux_var='_aux'):
- """
- functions like _serialize(), _unserialize(), and _unpack() sometimes need additional parameters:
- E.g. in order to unpack switch, extra parameters might be needed to evaluate the switch
- expression. This function tries to resolve all fields within a structure, and returns the
- unresolved fields as the list of external parameters.
- """
- def add_param(params, param):
- if param not in params:
- params.append(param)
-
- # collect all fields into param_fields
- param_fields = []
- wire_fields = []
-
- for field in self.fields:
- if field.visible:
- # the field should appear as a parameter in the function call
- param_fields.append(field)
- if field.wire and not field.auto:
- if field.type.fixed_size() and not self.is_switch:
- # field in the xcb_out structure
- wire_fields.append(field)
- # fields like 'pad0' are skipped!
-
- # in case of switch, parameters always contain any fields referenced in the switch expr
- # we do not need any variable size fields here, as the switch data type contains both
- # fixed and variable size fields
- if self.is_switch:
- param_fields = get_expr_fields(self)
-
- # _serialize()/_unserialize()/_unpack() function parameters
- # note: don't use set() for params, it is unsorted
- params = []
-
- # 1. the parameter for the void * buffer
- if 'serialize' == context:
- params.append(('void', '**', buffer_var))
- elif context in ('unserialize', 'unpack', 'sizeof'):
- params.append(('const void', '*', buffer_var))
-
- # 2. any expr fields that cannot be resolved within self and descendants
- unresolved_fields = resolve_expr_fields(self)
- for f in unresolved_fields:
- add_param(params, (f.c_field_type, '', f.c_field_name))
-
- # 3. param_fields contain the fields necessary to evaluate the switch expr or any other fields
- # that do not appear in the data type struct
- for p in param_fields:
- if self.is_switch:
- typespec = p.c_field_const_type
- pointerspec = p.c_pointer
- add_param(params, (typespec, pointerspec, p.c_field_name))
- else:
- if p.visible and not p.wire and not p.auto:
- typespec = p.c_field_type
- pointerspec = ''
- add_param(params, (typespec, pointerspec, p.c_field_name))
-
- # 4. aux argument
- if 'serialize' == context:
- add_param(params, ('const %s' % self.c_type, '*', aux_var))
- elif 'unserialize' == context:
- add_param(params, ('%s' % self.c_type, '**', aux_var))
- elif 'unpack' == context:
- add_param(params, ('%s' % self.c_type, '*', aux_var))
-
- # 5. switch contains all variable size fields as struct members
- # for other data types though, these have to be supplied separately
- # this is important for the special case of intermixed fixed and
- # variable size fields
- if not self.is_switch and 'serialize' == context:
- for p in param_fields:
- if not p.type.fixed_size():
- add_param(params, (p.c_field_const_type, '*', p.c_field_name))
-
- return (param_fields, wire_fields, params)
-# get_serialize_params()
-
-def _c_serialize_helper_insert_padding(context, code_lines, space, postpone):
- code_lines.append('%s /* insert padding */' % space)
- code_lines.append('%s xcb_pad = -xcb_block_len & 3;' % space)
-# code_lines.append('%s printf("automatically inserting padding: %%%%d\\n", xcb_pad);' % space)
- code_lines.append('%s xcb_buffer_len += xcb_block_len + xcb_pad;' % space)
-
- if not postpone:
- code_lines.append('%s if (0 != xcb_pad) {' % space)
-
- if 'serialize' == context:
- code_lines.append('%s xcb_parts[xcb_parts_idx].iov_base = xcb_pad0;' % space)
- code_lines.append('%s xcb_parts[xcb_parts_idx].iov_len = xcb_pad;' % space)
- code_lines.append('%s xcb_parts_idx++;' % space)
- elif context in ('unserialize', 'unpack', 'sizeof'):
- code_lines.append('%s xcb_tmp += xcb_pad;' % space)
-
- code_lines.append('%s xcb_pad = 0;' % space)
- code_lines.append('%s }' % space)
-
- code_lines.append('%s xcb_block_len = 0;' % space)
-
- # keep tracking of xcb_parts entries for serialize
- return 1
-# _c_serialize_helper_insert_padding()
-
-def _c_serialize_helper_switch(context, self, complex_name,
- code_lines, temp_vars,
- space, prefix):
- count = 0
- switch_expr = _c_accessor_get_expr(self.expr, None)
-
- for b in self.bitcases:
- bitcase_expr = _c_accessor_get_expr(b.type.expr, None)
- code_lines.append(' if(%s & %s) {' % (switch_expr, bitcase_expr))
-# code_lines.append(' printf("switch %s: entering bitcase section %s (mask=%%%%d)...\\n", %s);' %
-# (self.name[-1], b.type.name[-1], bitcase_expr))
- b_prefix = prefix
- if b.type.has_name:
- b_prefix = prefix + [(b.c_field_name, '.', b.type)]
-
- count += _c_serialize_helper_fields(context, b.type,
- code_lines, temp_vars,
- "%s " % space,
- b_prefix,
- is_bitcase = True)
- code_lines.append(' }')
-
-# if 'serialize' == context:
-# count += _c_serialize_helper_insert_padding(context, code_lines, space, False)
-# elif context in ('unserialize', 'unpack', 'sizeof'):
-# # padding
-# code_lines.append('%s xcb_pad = -xcb_block_len & 3;' % space)
-# code_lines.append('%s xcb_buffer_len += xcb_block_len + xcb_pad;' % space)
-
- return count
-# _c_serialize_helper_switch
-
-def _c_serialize_helper_switch_field(context, self, field, c_switch_variable, prefix):
- """
- handle switch by calling _serialize() or _unpack(), depending on context
- """
- # switch is handled by this function as a special case
- param_fields, wire_fields, params = get_serialize_params(context, self)
- field_mapping = _c_helper_field_mapping(self, prefix)
- prefix_str = _c_helper_absolute_name(prefix)
-
- # find the parameters that need to be passed to _serialize()/_unpack():
- # all switch expr fields must be given as parameters
- args = get_expr_fields(field.type)
- # length fields for variable size types in switch, normally only some of need
- # need to be passed as parameters
- switch_len_fields = resolve_expr_fields(field.type)
-
- # a switch field at this point _must_ be a bitcase field
- # we require that bitcases are "self-contiguous"
- bitcase_unresolved = resolve_expr_fields(self)
- if len(bitcase_unresolved) != 0:
- raise Exception('unresolved fields within bitcase is not supported at this point')
-
- # get the C names for the parameters
- c_field_names = ''
- for a in switch_len_fields:
- c_field_names += "%s, " % field_mapping[a.c_field_name][0]
- for a in args:
- c_field_names += "%s, " % field_mapping[a.c_field_name][0]
-
- # call _serialize()/_unpack() to determine the actual size
- if 'serialize' == context:
- length = "%s(&%s, %s&%s%s)" % (field.type.c_serialize_name, c_switch_variable,
- c_field_names, prefix_str, field.c_field_name)
- elif context in ('unserialize', 'unpack'):
- length = "%s(xcb_tmp, %s&%s%s)" % (field.type.c_unpack_name,
- c_field_names, prefix_str, field.c_field_name)
-
- return length
-# _c_serialize_helper_switch_field()
-
-def _c_serialize_helper_list_field(context, self, field,
- code_lines, temp_vars,
- space, prefix):
- """
- helper function to cope with lists of variable length
- """
- expr = field.type.expr
- prefix_str = _c_helper_absolute_name(prefix)
- param_fields, wire_fields, params = get_serialize_params('sizeof', self)
- param_names = [p[2] for p in params]
-
- expr_fields_names = [f.field_name for f in get_expr_fields(field.type)]
- resolved = filter(lambda x: x in param_names, expr_fields_names)
- unresolved = filter(lambda x: x not in param_names, expr_fields_names)
-
- field_mapping = {}
- for r in resolved:
- field_mapping[r] = (r, None)
-
- if len(unresolved)>0:
- tmp_prefix = prefix
- if len(tmp_prefix)==0:
- raise Exception("found an empty prefix while resolving expr field names for list %s",
- field.c_field_name)
-
- field_mapping.update(_c_helper_resolve_field_names(prefix))
- resolved += filter(lambda x: x in field_mapping, unresolved)
- unresolved = filter(lambda x: x not in field_mapping, unresolved)
- if len(unresolved)>0:
- raise Exception('could not resolve the length fields required for list %s' % field.c_field_name)
-
- list_length = _c_accessor_get_expr(expr, field_mapping)
-
- # default: list with fixed size elements
- length = '%s * sizeof(%s)' % (list_length, field.type.member.c_wiretype)
-
- # list with variable-sized elements
- if not field.type.member.fixed_size():
- length = ''
- if context in ('unserialize', 'sizeof', 'unpack'):
- int_i = ' unsigned int i;'
- xcb_tmp_len = ' unsigned int xcb_tmp_len;'
- if int_i not in temp_vars:
- temp_vars.append(int_i)
- if xcb_tmp_len not in temp_vars:
- temp_vars.append(xcb_tmp_len)
- # loop over all list elements and call sizeof repeatedly
- # this should be a bit faster than using the iterators
- code_lines.append("%s for(i=0; i<%s; i++) {" % (space, list_length))
- code_lines.append("%s xcb_tmp_len = %s(xcb_tmp);" %
- (space, field.type.c_sizeof_name))
- code_lines.append("%s xcb_block_len += xcb_tmp_len;" % space)
- code_lines.append("%s xcb_tmp += xcb_tmp_len;" % space)
- code_lines.append("%s }" % space)
-
- elif 'serialize' == context:
- code_lines.append('%s xcb_parts[xcb_parts_idx].iov_len = 0;' % space)
- code_lines.append('%s xcb_tmp = (char *) %s%s;' % (space, prefix_str, field.c_field_name))
- code_lines.append('%s for(i=0; i<%s; i++) { ' % (space, list_length))
- code_lines.append('%s xcb_block_len = %s(xcb_tmp);' % (space, field.type.c_sizeof_name))
- code_lines.append('%s xcb_parts[xcb_parts_idx].iov_len += xcb_block_len;' % space)
- code_lines.append('%s }' % space)
- code_lines.append('%s xcb_block_len = xcb_parts[xcb_parts_idx].iov_len;' % space)
-
- return length
-# _c_serialize_helper_list_field()
-
-def _c_serialize_helper_fields_fixed_size(context, self, field,
- code_lines, temp_vars,
- space, prefix):
- # keep the C code a bit more readable by giving the field name
- if not self.is_bitcase:
- code_lines.append('%s /* %s.%s */' % (space, self.c_type, field.c_field_name))
- else:
- scoped_name = [p[2].c_type if idx==0 else p[0] for idx, p in enumerate(prefix)]
- typename = reduce(lambda x,y: "%s.%s" % (x, y), scoped_name)
- code_lines.append('%s /* %s.%s */' % (space, typename, field.c_field_name))
-
- abs_field_name = _c_helper_absolute_name(prefix, field)
- # default for simple cases: call sizeof()
- length = "sizeof(%s)" % field.c_field_type
-
- if context in ('unserialize', 'unpack', 'sizeof'):
- # default: simple cast
- value = ' %s = *(%s *)xcb_tmp;' % (abs_field_name, field.c_field_type)
-
- # padding - we could probably just ignore it
- if field.type.is_pad and field.type.nmemb > 1:
- value = ''
- for i in range(field.type.nmemb):
- code_lines.append('%s %s[%d] = *(%s *)xcb_tmp;' %
- (space, abs_field_name, i, field.c_field_type))
- # total padding = sizeof(pad0) * nmemb
- length += " * %d" % field.type.nmemb
-
- if field.type.is_list:
- # no such case in the protocol, cannot be tested and therefore ignored for now
- raise Exception('list with fixed number of elemens unhandled in _unserialize()')
-
- elif 'serialize' == context:
- value = ' xcb_parts[xcb_parts_idx].iov_base = (char *) '
-
- if field.type.is_expr:
- # need to register a temporary variable for the expression in case we know its type
- if field.type.c_type is None:
- raise Exception("type for field '%s' (expression '%s') unkown" %
- (field.field_name, _c_accessor_get_expr(field.type.expr)))
-
- temp_vars.append(' %s xcb_expr_%s = %s;' % (field.type.c_type, _cpp(field.field_name),
- _c_accessor_get_expr(field.type.expr, prefix)))
- value += "&xcb_expr_%s;" % _cpp(field.field_name)
-
- elif field.type.is_pad:
- if field.type.nmemb == 1:
- value += "&xcb_pad;"
- else:
- # we could also set it to 0, see definition of xcb_send_request()
- value = ' xcb_parts[xcb_parts_idx].iov_base = xcb_pad0;'
- length += "*%d" % field.type.nmemb
-
- else:
- # non-list type with fixed size
- if field.type.nmemb == 1:
- value += "&%s;" % (abs_field_name)
-
- # list with nmemb (fixed size) elements
- else:
- value += '%s;' % (abs_field_name)
- length = '%d' % field.type.nmemb
-
- return (value, length)
-# _c_serialize_helper_fields_fixed_size()
-
-def _c_serialize_helper_fields_variable_size(context, self, field,
- code_lines, temp_vars,
- space, prefix):
- prefix_str = _c_helper_absolute_name(prefix)
-
- if context in ('unserialize', 'unpack', 'sizeof'):
- value = ''
- var_field_name = 'xcb_tmp'
-
- # special case: intermixed fixed and variable size fields
- if self.var_followed_by_fixed_fields and 'unserialize' == context:
- value = ' %s = (%s *)xcb_tmp;' % (field.c_field_name, field.c_field_type)
- temp_vars.append(' %s *%s;' % (field.type.c_type, field.c_field_name))
- # special case: switch
- if 'unpack' == context:
- value = ' %s%s = (%s *)xcb_tmp;' % (prefix_str, field.c_field_name, field.c_field_type)
-
- elif 'serialize' == context:
- # variable size fields appear as parameters to _serialize() if the
- # 'toplevel' container is not a switch
- prefix_string = prefix_str if prefix[0][2].is_switch else ''
- var_field_name = "%s%s" % (prefix_string, field.c_field_name)
- value = ' xcb_parts[xcb_parts_idx].iov_base = (char *) %s;' % var_field_name
-
- length = ''
-
- code_lines.append('%s /* %s */' % (space, field.c_field_name))
-
- if field.type.is_list:
- if value != '':
- # in any context, list is already a pointer, so the default assignment is ok
- code_lines.append("%s%s" % (space, value))
- value = ''
- length = _c_serialize_helper_list_field(context, self, field,
- code_lines, temp_vars,
- space, prefix)
-
- elif field.type.is_switch:
- value = ''
- if context == 'serialize':
- # the _serialize() function allocates the correct amount memory if given a NULL pointer
- value = ' xcb_parts[xcb_parts_idx].iov_base = (char *)0;'
- length = _c_serialize_helper_switch_field(context, self, field,
- 'xcb_parts[xcb_parts_idx].iov_base',
- prefix)
-
- else:
- # in all remaining special cases - call _sizeof()
- length = "%s(%s)" % (field.type.c_sizeof_name, var_field_name)
-
- return (value, length)
-# _c_serialize_helper_fields_variable_size
-
-def _c_serialize_helper_fields(context, self,
- code_lines, temp_vars,
- space, prefix, is_bitcase):
- count = 0
- need_padding = False
- prev_field_was_variable = False
-
- for field in self.fields:
- if not field.visible:
- if not ((field.wire and not field.auto) or 'unserialize' == context):
- continue
-
- # switch/bitcase: fixed size fields must be considered explicitly
- if field.type.fixed_size():
- if self.is_bitcase or self.var_followed_by_fixed_fields:
- if prev_field_was_variable and need_padding:
- # insert padding
-# count += _c_serialize_helper_insert_padding(context, code_lines, space,
-# self.var_followed_by_fixed_fields)
- prev_field_was_variable = False
-
- # prefix for fixed size fields
- fixed_prefix = prefix
-
- value, length = _c_serialize_helper_fields_fixed_size(context, self, field,
- code_lines, temp_vars,
- space, fixed_prefix)
- else:
- continue
-
- # fields with variable size
- else:
- # switch/bitcase: always calculate padding before and after variable sized fields
- if need_padding or is_bitcase:
- count += _c_serialize_helper_insert_padding(context, code_lines, space,
- self.var_followed_by_fixed_fields)
-
- value, length = _c_serialize_helper_fields_variable_size(context, self, field,
- code_lines, temp_vars,
- space, prefix)
- prev_field_was_variable = True
-
- # save (un)serialization C code
- if '' != value:
- code_lines.append('%s%s' % (space, value))
-
- if field.type.fixed_size():
- if is_bitcase or self.var_followed_by_fixed_fields:
- # keep track of (un)serialized object's size
- code_lines.append('%s xcb_block_len += %s;' % (space, length))
- if context in ('unserialize', 'unpack', 'sizeof'):
- code_lines.append('%s xcb_tmp += %s;' % (space, length))
- else:
- # variable size objects or bitcase:
- # value & length might have been inserted earlier for special cases
- if '' != length:
- # special case: intermixed fixed and variable size fields
- if (not field.type.fixed_size() and
- self.var_followed_by_fixed_fields and 'unserialize' == context):
- temp_vars.append(' int %s_len;' % field.c_field_name)
- code_lines.append('%s %s_len = %s;' % (space, field.c_field_name, length))
- code_lines.append('%s xcb_block_len += %s_len;' % (space, field.c_field_name))
- code_lines.append('%s xcb_tmp += %s_len;' % (space, field.c_field_name))
- else:
- code_lines.append('%s xcb_block_len += %s;' % (space, length))
- # increase pointer into the byte stream accordingly
- if context in ('unserialize', 'sizeof', 'unpack'):
- code_lines.append('%s xcb_tmp += xcb_block_len;' % space)
-
- if 'serialize' == context:
- if '' != length:
- code_lines.append('%s xcb_parts[xcb_parts_idx].iov_len = %s;' % (space, length))
- code_lines.append('%s xcb_parts_idx++;' % space)
- count += 1
-
- need_padding = True
- if self.var_followed_by_fixed_fields:
- need_padding = False
-
- return count
-# _c_serialize_helper_fields()
-
-def _c_serialize_helper(context, complex_type,
- code_lines, temp_vars,
- space='', prefix=[]):
- # count tracks the number of fields to serialize
- count = 0
-
- if hasattr(complex_type, 'type'):
- self = complex_type.type
- complex_name = complex_type.name
- else:
- self = complex_type
- if self.var_followed_by_fixed_fields and 'unserialize' == context:
- complex_name = 'xcb_out'
- else:
- complex_name = '_aux'
-
- # special case: switch is serialized by evaluating each bitcase separately
- if self.is_switch:
- count += _c_serialize_helper_switch(context, self, complex_name,
- code_lines, temp_vars,
- space, prefix)
-
- # all other data types can be evaluated one field a time
- else:
- # unserialize & fixed size fields: simply cast the buffer to the respective xcb_out type
- if context in ('unserialize', 'unpack', 'sizeof') and not self.var_followed_by_fixed_fields:
- code_lines.append('%s xcb_block_len += sizeof(%s);' % (space, self.c_type))
- code_lines.append('%s xcb_tmp += xcb_block_len;' % space)
- # probably not needed
- #_c_serialize_helper_insert_padding(context, code_lines, space, False)
-
- count += _c_serialize_helper_fields(context, self,
- code_lines, temp_vars,
- space, prefix, False)
- # "final padding"
- count += _c_serialize_helper_insert_padding(context, code_lines, space, False)
-
- return count
-# _c_serialize_helper()
-
-def _c_serialize(context, self):
- """
- depending on the context variable, generate _serialize(), _unserialize(), _unpack(), or _sizeof()
- for the ComplexType variable self
- """
- _h_setlevel(1)
- _c_setlevel(1)
-
- _hc('')
- # _serialize() returns the buffer size
- _hc('int')
-
- if self.is_switch and 'unserialize' == context:
- context = 'unpack'
-
- cases = { 'serialize' : self.c_serialize_name,
- 'unserialize' : self.c_unserialize_name,
- 'unpack' : self.c_unpack_name,
- 'sizeof' : self.c_sizeof_name }
- func_name = cases[context]
-
- param_fields, wire_fields, params = get_serialize_params(context, self)
- variable_size_fields = 0
- # maximum space required for type definition of function arguments
- maxtypelen = 0
-
- # determine N(variable_fields)
- for field in param_fields:
- # if self.is_switch, treat all fields as if they are variable sized
- if not field.type.fixed_size() or self.is_switch:
- variable_size_fields += 1
- # determine maxtypelen
- for p in params:
- maxtypelen = max(maxtypelen, len(p[0]) + len(p[1]))
-
- # write to .c/.h
- indent = ' '*(len(func_name)+2)
- param_str = []
- for p in params:
- typespec, pointerspec, field_name = p
- spacing = ' '*(maxtypelen-len(typespec)-len(pointerspec))
- param_str.append("%s%s%s %s%s /**< */" % (indent, typespec, spacing, pointerspec, field_name))
- # insert function name
- param_str[0] = "%s (%s" % (func_name, param_str[0].strip())
- param_str = map(lambda x: "%s," % x, param_str)
- for s in param_str[:-1]:
- _hc(s)
- _h("%s);" % param_str[-1].rstrip(','))
- _c("%s)" % param_str[-1].rstrip(','))
- _c('{')
-
- code_lines = []
- temp_vars = []
- prefix = []
-
- if 'serialize' == context:
- if not self.is_switch and not self.var_followed_by_fixed_fields:
- _c(' %s *xcb_out = *_buffer;', self.c_type)
- _c(' unsigned int xcb_out_pad = -sizeof(%s) & 3;', self.c_type)
- _c(' unsigned int xcb_buffer_len = sizeof(%s) + xcb_out_pad;', self.c_type)
- else:
- _c(' char *xcb_out = *_buffer;')
- _c(' unsigned int xcb_buffer_len = 0;')
- prefix = [('_aux', '->', self)]
- aux_ptr = 'xcb_out'
-
- elif context in ('unserialize', 'unpack'):
- _c(' char *xcb_tmp = (char *)_buffer;')
- if not self.is_switch:
- if not self.var_followed_by_fixed_fields:
- _c(' const %s *_aux = (%s *)_buffer;', self.c_type, self.c_type)
- prefix = [('_aux', '->', self)]
- else:
- _c(' %s xcb_out;', self.c_type)
- prefix = [('xcb_out', '.', self)]
- else:
- aux_var = '_aux' # default for unpack: single pointer
- # note: unserialize not generated for switch
- if 'unserialize' == context:
- aux_var = '(*_aux)' # unserialize: double pointer (!)
- prefix = [(aux_var, '->', self)]
- aux_ptr = '*_aux'
- _c(' unsigned int xcb_buffer_len = 0;')
- _c(' unsigned int xcb_block_len = 0;')
- _c(' unsigned int xcb_pad = 0;')
-
- elif 'sizeof' == context:
- param_names = [p[2] for p in params]
- if self.is_switch:
- # switch: call _unpack()
- _c(' %s _aux;', self.c_type)
- _c(' return %s(%s, &_aux);', self.c_unpack_name, reduce(lambda x,y: "%s, %s" % (x, y), param_names))
- _c('}')
- return
- elif self.var_followed_by_fixed_fields:
- # special case: call _unserialize()
- _c(' return %s(%s, NULL);', self.c_unserialize_name, reduce(lambda x,y: "%s, %s" % (x, y), param_names))
- _c('}')
- return
- else:
- _c(' char *xcb_tmp = (char *)_buffer;')
- prefix = [('_aux', '->', self)]
-
- count = _c_serialize_helper(context, self, code_lines, temp_vars, prefix=prefix)
- # update variable size fields (only important for context=='serialize'
- variable_size_fields = count
- if 'serialize' == context:
- temp_vars.append(' unsigned int xcb_pad = 0;')
- temp_vars.append(' char xcb_pad0[3] = {0, 0, 0};')
- temp_vars.append(' struct iovec xcb_parts[%d];' % count)
- temp_vars.append(' unsigned int xcb_parts_idx = 0;')
- temp_vars.append(' unsigned int xcb_block_len = 0;')
- temp_vars.append(' unsigned int i;')
- temp_vars.append(' char *xcb_tmp;')
- elif 'sizeof' == context:
- # neither switch nor intermixed fixed and variable size fields:
- # evaluate parameters directly
- if not (self.is_switch or self.var_followed_by_fixed_fields):
-
- # look if we have to declare an '_aux' variable at all
- if len(filter(lambda x: x.find('_aux')!=-1, code_lines))>0:
- if not self.var_followed_by_fixed_fields:
- _c(' const %s *_aux = (%s *)_buffer;', self.c_type, self.c_type)
- else:
- _c(' %s *_aux = malloc(sizeof(%s));', self.c_type, self.c_type)
-
- _c(' unsigned int xcb_buffer_len = 0;')
- _c(' unsigned int xcb_block_len = 0;')
- _c(' unsigned int xcb_pad = 0;')
-
- _c('')
- for t in temp_vars:
- _c(t)
- _c('')
- for l in code_lines:
- _c(l)
-
- # variable sized fields have been collected, now
- # allocate memory and copy everything into a continuous memory area
- # note: this is not necessary in case of unpack
- if context in ('serialize', 'unserialize'):
- # unserialize: check for sizeof-only invocation
- if 'unserialize' == context:
- _c('')
- _c(' if (NULL == _aux)')
- _c(' return xcb_buffer_len;')
-
- _c('')
- _c(' if (NULL == %s) {', aux_ptr)
- _c(' /* allocate memory */')
- _c(' %s = malloc(xcb_buffer_len);', aux_ptr)
- if 'serialize' == context:
- _c(' *_buffer = xcb_out;')
- _c(' }')
- _c('')
-
- # serialize: handle variable size fields in a loop
- if 'serialize' == context:
- if not self.is_switch and not self.var_followed_by_fixed_fields:
- if len(wire_fields)>0:
- _c(' *xcb_out = *_aux;')
- # copy variable size fields into the buffer
- if variable_size_fields > 0:
- # xcb_out padding
- if not self.is_switch and not self.var_followed_by_fixed_fields:
- _c(' xcb_tmp = (char*)++xcb_out;')
- _c(' xcb_tmp += xcb_out_pad;')
- else:
- _c(' xcb_tmp = xcb_out;')
-
- # variable sized fields
- _c(' for(i=0; i<xcb_parts_idx; i++) {')
- _c(' if (0 != xcb_parts[i].iov_base && 0 != xcb_parts[i].iov_len)')
- _c(' memcpy(xcb_tmp, xcb_parts[i].iov_base, xcb_parts[i].iov_len);')
- _c(' if (0 != xcb_parts[i].iov_len)')
- _c(' xcb_tmp += xcb_parts[i].iov_len;')
- _c(' }')
-
- # unserialize: assign variable size fields individually
- if 'unserialize' == context:
- _c(' xcb_tmp = ((char *)*_aux)+xcb_buffer_len;')
- param_fields.reverse()
- for field in param_fields:
- if not field.type.fixed_size():
- _c(' xcb_tmp -= %s_len;', field.c_field_name)
- _c(' memmove(xcb_tmp, %s, %s_len);', field.c_field_name, field.c_field_name)
- _c(' *%s = xcb_out;', aux_ptr)
-
- _c('')
- _c(' return xcb_buffer_len;')
- _c('}')
-# _c_serialize()
-
-def _c_iterator_get_end(field, accum):
- '''
- Figures out what C code is needed to find the end of a variable-length structure field.
- For nested structures, recurses into its last variable-sized field.
- For lists, calls the end function
- '''
- if field.type.is_container:
- accum = field.c_accessor_name + '(' + accum + ')'
- return _c_iterator_get_end(field.type.last_varsized_field, accum)
- if field.type.is_list:
- # XXX we can always use the first way
- if field.type.member.is_simple:
- return field.c_end_name + '(' + accum + ')'
- else:
- return field.type.member.c_end_name + '(' + field.c_iterator_name + '(' + accum + '))'
-
-def _c_iterator(self, name):
- '''
- Declares the iterator structure and next/end functions for a given type.
- '''
- _h_setlevel(0)
- _h('')
- _h('/**')
- _h(' * @brief %s', self.c_iterator_type)
- _h(' **/')
- _h('typedef struct %s {', self.c_iterator_type)
- _h(' %s *data; /**< */', self.c_type)
- _h(' int%s rem; /**< */', ' ' * (len(self.c_type) - 2))
- _h(' int%s index; /**< */', ' ' * (len(self.c_type) - 2))
- _h('} %s;', self.c_iterator_type)
-
- _h_setlevel(1)
- _c_setlevel(1)
- _h('')
- _h('/**')
- _h(' * Get the next element of the iterator')
- _h(' * @param i Pointer to a %s', self.c_iterator_type)
- _h(' *')
- _h(' * Get the next element in the iterator. The member rem is')
- _h(' * decreased by one. The member data points to the next')
- _h(' * element. The member index is increased by sizeof(%s)', self.c_type)
- _h(' */')
- _c('')
- _hc('')
- _hc('/*****************************************************************************')
- _hc(' **')
- _hc(' ** void %s', self.c_next_name)
- _hc(' ** ')
- _hc(' ** @param %s *i', self.c_iterator_type)
- _hc(' ** @returns void')
- _hc(' **')
- _hc(' *****************************************************************************/')
- _hc(' ')
- _hc('void')
- _h('%s (%s *i /**< */);', self.c_next_name, self.c_iterator_type)
- _c('%s (%s *i /**< */)', self.c_next_name, self.c_iterator_type)
- _c('{')
-
- if not self.fixed_size():
- _c(' %s *R = i->data;', self.c_type)
-
- if self.is_union:
- # FIXME - how to determine the size of a variable size union??
- _c(' /* FIXME - determine the size of the union %s */', self.c_type)
- else:
- if self.need_sizeof:
- _c(' xcb_generic_iterator_t child;')
- _c(' child.data = (%s *)(((char *)R) + %s(R));',
- self.c_type, self.c_sizeof_name)
- _c(' i->index = (char *) child.data - (char *) i->data;')
- else:
- _c(' xcb_generic_iterator_t child = %s;', _c_iterator_get_end(self.last_varsized_field, 'R'))
- _c(' i->index = child.index;')
- _c(' --i->rem;')
- _c(' i->data = (%s *) child.data;', self.c_type)
-
- else:
- _c(' --i->rem;')
- _c(' ++i->data;')
- _c(' i->index += sizeof(%s);', self.c_type)
-
- _c('}')
-
- _h('')
- _h('/**')
- _h(' * Return the iterator pointing to the last element')
- _h(' * @param i An %s', self.c_iterator_type)
- _h(' * @return The iterator pointing to the last element')
- _h(' *')
- _h(' * Set the current element in the iterator to the last element.')
- _h(' * The member rem is set to 0. The member data points to the')
- _h(' * last element.')
- _h(' */')
- _c('')
- _hc('')
- _hc('/*****************************************************************************')
- _hc(' **')
- _hc(' ** xcb_generic_iterator_t %s', self.c_end_name)
- _hc(' ** ')
- _hc(' ** @param %s i', self.c_iterator_type)
- _hc(' ** @returns xcb_generic_iterator_t')
- _hc(' **')
- _hc(' *****************************************************************************/')
- _hc(' ')
- _hc('xcb_generic_iterator_t')
- _h('%s (%s i /**< */);', self.c_end_name, self.c_iterator_type)
- _c('%s (%s i /**< */)', self.c_end_name, self.c_iterator_type)
- _c('{')
- _c(' xcb_generic_iterator_t ret;')
-
- if self.fixed_size():
- _c(' ret.data = i.data + i.rem;')
- _c(' ret.index = i.index + ((char *) ret.data - (char *) i.data);')
- _c(' ret.rem = 0;')
- else:
- _c(' while(i.rem > 0)')
- _c(' %s(&i);', self.c_next_name)
- _c(' ret.data = i.data;')
- _c(' ret.rem = i.rem;')
- _c(' ret.index = i.index;')
-
- _c(' return ret;')
- _c('}')
-
-def _c_accessor_get_length(expr, field_mapping=None):
- '''
- Figures out what C code is needed to get a length field.
- The field_mapping parameter can be used to change the absolute name of a length field.
- For fields that follow a variable-length field, use the accessor.
- Otherwise, just reference the structure field directly.
- '''
-
- lenfield_name = expr.lenfield_name
- if lenfield_name is not None:
- if field_mapping is not None:
- lenfield_name = field_mapping[lenfield_name][0]
-
- if expr.lenfield is not None and expr.lenfield.prev_varsized_field is not None:
- # special case: variable and fixed size fields are intermixed
- # if the lenfield is among the fixed size fields, there is no need
- # to call a special accessor function like <expr.lenfield.c_accessor_name + '(' + prefix + ')'>
- return field_mapping(expr.lenfield_name)
- elif expr.lenfield_name is not None:
- return lenfield_name
- else:
- return str(expr.nmemb)
-
-def _c_accessor_get_expr(expr, field_mapping):
- '''
- Figures out what C code is needed to get the length of a list field.
- The field_mapping parameter can be used to change the absolute name of a length field.
- Recurses for math operations.
- Returns bitcount for value-mask fields.
- Otherwise, uses the value of the length field.
- '''
- lenexp = _c_accessor_get_length(expr, field_mapping)
-
- if expr.op == '~':
- return '(' + '~' + _c_accessor_get_expr(expr.rhs, field_mapping) + ')'
- elif expr.op == 'popcount':
- return 'xcb_popcount(' + _c_accessor_get_expr(expr.rhs, field_mapping) + ')'
- elif expr.op == 'enumref':
- enum_name = expr.lenfield_type.name
- constant_name = expr.lenfield_name
- c_name = _n(enum_name + (constant_name,)).upper()
- return c_name
- elif expr.op == 'sumof':
- # locate the referenced list object
- list_obj = expr.lenfield_type
- field = None
- for f in expr.lenfield_parent.fields:
- if f.field_name == expr.lenfield_name:
- field = f
- break
-
- if field is None:
- raise Exception("list field '%s' referenced by sumof not found" % expr.lenfield_name)
- list_name = field_mapping[field.c_field_name][0]
- c_length_func = "%s(%s)" % (field.c_length_name, list_name)
- # note: xcb_sumof() has only been defined for integers
- c_length_func = _c_accessor_get_expr(field.type.expr, field_mapping)
- return 'xcb_sumof(%s, %s)' % (list_name, c_length_func)
- elif expr.op != None:
- return ('(' + _c_accessor_get_expr(expr.lhs, field_mapping) +
- ' ' + expr.op + ' ' +
- _c_accessor_get_expr(expr.rhs, field_mapping) + ')')
- elif expr.bitfield:
- return 'xcb_popcount(' + lenexp + ')'
- else:
- return lenexp
-
-def _c_accessors_field(self, field):
- '''
- Declares the accessor functions for a non-list field that follows a variable-length field.
- '''
- c_type = self.c_type
-
- # special case: switch
- switch_obj = self if self.is_switch else None
- if self.is_bitcase:
- switch_obj = self.parents[-1]
- if switch_obj is not None:
- c_type = switch_obj.c_type
-
- if field.type.is_simple:
- _hc('')
- _hc('')
- _hc('/*****************************************************************************')
- _hc(' ** ')
- _hc(' ** %s %s', field.c_field_type, field.c_accessor_name)
- _hc(' ** ')
- _hc(' ** @param const %s *R', c_type)
- _hc(' ** @returns %s', field.c_field_type)
- _hc(' **')
- _hc(' *****************************************************************************/')
- _hc(' ')
- _hc('%s', field.c_field_type)
- _h('%s (const %s *R /**< */);', field.c_accessor_name, c_type)
- _c('%s (const %s *R /**< */)', field.c_accessor_name, c_type)
- _c('{')
- if field.prev_varsized_field is None:
- _c(' return (%s *) (R + 1);', field.c_field_type)
- else:
- _c(' xcb_generic_iterator_t prev = %s;', _c_iterator_get_end(field.prev_varsized_field, 'R'))
- _c(' return * (%s *) ((char *) prev.data + XCB_TYPE_PAD(%s, prev.index) + %d);',
- field.c_field_type, field.first_field_after_varsized.type.c_type, field.prev_varsized_offset)
- _c('}')
- else:
- _hc('')
- _hc('')
- _hc('/*****************************************************************************')
- _hc(' **')
- _hc(' ** %s * %s', field.c_field_type, field.c_accessor_name)
- _hc(' ** ')
- _hc(' ** @param const %s *R', c_type)
- _hc(' ** @returns %s *', field.c_field_type)
- _hc(' **')
- _hc(' *****************************************************************************/')
- _hc(' ')
- if field.type.is_switch and switch_obj is None:
- return_type = 'void *'
- else:
- return_type = '%s *' % field.c_field_type
-
- _hc(return_type)
- _h('%s (const %s *R /**< */);', field.c_accessor_name, c_type)
- _c('%s (const %s *R /**< */)', field.c_accessor_name, c_type)
- _c('{')
- if field.prev_varsized_field is None:
- _c(' return (%s) (R + 1);', return_type)
- # note: the special case 'variable fields followed by fixed size fields'
- # is not of any consequence here, since the ordering gets
- # 'corrected' in the reply function
- else:
- _c(' xcb_generic_iterator_t prev = %s;', _c_iterator_get_end(field.prev_varsized_field, 'R'))
- _c(' return (%s) ((char *) prev.data + XCB_TYPE_PAD(%s, prev.index) + %d);',
- return_type, field.first_field_after_varsized.type.c_type, field.prev_varsized_offset)
- _c('}')
-
-
-def _c_accessors_list(self, field):
- '''
- Declares the accessor functions for a list field.
- Declares a direct-accessor function only if the list members are fixed size.
- Declares length and get-iterator functions always.
- '''
- list = field.type
- c_type = self.c_type
-
- # special case: switch
- # in case of switch, 2 params have to be supplied to certain accessor functions:
- # 1. the anchestor object (request or reply)
- # 2. the (anchestor) switch object
- # the reason is that switch is either a child of a request/reply or nested in another switch,
- # so whenever we need to access a length field, we might need to refer to some anchestor type
- switch_obj = self if self.is_switch else None
- if self.is_bitcase:
- switch_obj = self.parents[-1]
- if switch_obj is not None:
- c_type = switch_obj.c_type
-
- params = []
- fields = {}
- parents = self.parents if hasattr(self, 'parents') else [self]
- # 'R': parents[0] is always the 'toplevel' container type
- params.append(('const %s *R' % parents[0].c_type, parents[0]))
- fields.update(_c_helper_field_mapping(parents[0], [('R', '->', parents[0])], flat=True))
- # auxiliary object for 'R' parameters
- R_obj = parents[0]
-
- if switch_obj is not None:
- # now look where the fields are defined that are needed to evaluate
- # the switch expr, and store the parent objects in accessor_params and
- # the fields in switch_fields
-
- # 'S': name for the 'toplevel' switch
- toplevel_switch = parents[1]
- params.append(('const %s *S' % toplevel_switch.c_type, toplevel_switch))
- fields.update(_c_helper_field_mapping(toplevel_switch, [('S', '->', toplevel_switch)], flat=True))
-
- # initialize prefix for everything "below" S
- prefix_str = '/* %s */ S' % toplevel_switch.name[-1]
- prefix = [(prefix_str, '->', toplevel_switch)]
-
- # look for fields in the remaining containers
- for p in parents[2:] + [self]:
- # the separator between parent and child is always '.' here,
- # because of nested switch statements
- if not p.is_bitcase or (p.is_bitcase and p.has_name):
- prefix.append((p.name[-1], '.', p))
- fields.update(_c_helper_field_mapping(p, prefix, flat=True))
-
- # auxiliary object for 'S' parameter
- S_obj = parents[1]
-
- _h_setlevel(1)
- _c_setlevel(1)
- if list.member.fixed_size():
- idx = 1 if switch_obj is not None else 0
- _hc('')
- _hc('')
- _hc('/*****************************************************************************')
- _hc(' **')
- _hc(' ** %s * %s', field.c_field_type, field.c_accessor_name)
- _hc(' ** ')
- _hc(' ** @param %s', params[idx][0])
- _hc(' ** @returns %s *', field.c_field_type)
- _hc(' **')
- _hc(' *****************************************************************************/')
- _hc(' ')
- _hc('%s *', field.c_field_type)
-
- _h('%s (%s /**< */);', field.c_accessor_name, params[idx][0])
- _c('%s (%s /**< */)', field.c_accessor_name, params[idx][0])
-
- _c('{')
- if switch_obj is not None:
- _c(' return %s;', fields[field.c_field_name][0])
- elif field.prev_varsized_field is None:
- _c(' return (%s *) (R + 1);', field.c_field_type)
- else:
- _c(' xcb_generic_iterator_t prev = %s;', _c_iterator_get_end(field.prev_varsized_field, 'R'))
- _c(' return (%s *) ((char *) prev.data + XCB_TYPE_PAD(%s, prev.index) + %d);',
- field.c_field_type, field.first_field_after_varsized.type.c_type, field.prev_varsized_offset)
- _c('}')
-
- _hc('')
- _hc('')
- _hc('/*****************************************************************************')
- _hc(' **')
- _hc(' ** int %s', field.c_length_name)
- _hc(' ** ')
- _hc(' ** @param const %s *R', c_type)
- _hc(' ** @returns int')
- _hc(' **')
- _hc(' *****************************************************************************/')
- _hc(' ')
- _hc('int')
- if switch_obj is not None:
- _hc('%s (const %s *R /**< */,', field.c_length_name, R_obj.c_type)
- spacing = ' '*(len(field.c_length_name)+2)
- _h('%sconst %s *S /**< */);', spacing, S_obj.c_type)
- _c('%sconst %s *S /**< */)', spacing, S_obj.c_type)
- length = _c_accessor_get_expr(field.type.expr, fields)
- else:
- _h('%s (const %s *R /**< */);', field.c_length_name, c_type)
- _c('%s (const %s *R /**< */)', field.c_length_name, c_type)
- length = _c_accessor_get_expr(field.type.expr, fields)
- _c('{')
- _c(' return %s;', length)
- _c('}')
-
- if field.type.member.is_simple:
- _hc('')
- _hc('')
- _hc('/*****************************************************************************')
- _hc(' **')
- _hc(' ** xcb_generic_iterator_t %s', field.c_end_name)
- _hc(' ** ')
- _hc(' ** @param const %s *R', c_type)
- _hc(' ** @returns xcb_generic_iterator_t')
- _hc(' **')
- _hc(' *****************************************************************************/')
- _hc(' ')
- _hc('xcb_generic_iterator_t')
- if switch_obj is not None:
- _hc('%s (const %s *R /**< */,', field.c_end_name, R_obj.c_type)
- spacing = ' '*(len(field.c_end_name)+2)
- _h('%sconst %s *S /**< */);', spacing, S_obj.c_type)
- _c('%sconst %s *S /**< */)', spacing, S_obj.c_type)
- else:
- _h('%s (const %s *R /**< */);', field.c_end_name, c_type)
- _c('%s (const %s *R /**< */)', field.c_end_name, c_type)
- _c('{')
- _c(' xcb_generic_iterator_t i;')
-
- param = 'R' if switch_obj is None else 'S'
- if switch_obj is not None:
- _c(' i.data = %s + %s;', fields[field.c_field_name][0],
- _c_accessor_get_expr(field.type.expr, fields))
- elif field.prev_varsized_field == None:
- _c(' i.data = ((%s *) (R + 1)) + (%s);', field.type.c_wiretype,
- _c_accessor_get_expr(field.type.expr, fields))
- else:
- _c(' xcb_generic_iterator_t child = %s;',
- _c_iterator_get_end(field.prev_varsized_field, 'R'))
- _c(' i.data = ((%s *) child.data) + (%s);', field.type.c_wiretype,
- _c_accessor_get_expr(field.type.expr, fields))
-
- _c(' i.rem = 0;')
- _c(' i.index = (char *) i.data - (char *) %s;', param)
- _c(' return i;')
- _c('}')
-
- else:
- _hc('')
- _hc('')
- _hc('/*****************************************************************************')
- _hc(' **')
- _hc(' ** %s %s', field.c_iterator_type, field.c_iterator_name)
- _hc(' ** ')
- _hc(' ** @param const %s *R', c_type)
- _hc(' ** @returns %s', field.c_iterator_type)
- _hc(' **')
- _hc(' *****************************************************************************/')
- _hc(' ')
-
- _hc('%s', field.c_iterator_type)
- if switch_obj is not None:
- _hc('%s (const %s *R /**< */,', field.c_iterator_name, R_obj.c_type)
- spacing = ' '*(len(field.c_iterator_name)+2)
- _h('%sconst %s *S /**< */);', spacing, S_obj.c_type)
- _c('%sconst %s *S /**< */)', spacing, S_obj.c_type)
- else:
- _h('%s (const %s *R /**< */);', field.c_iterator_name, c_type)
- _c('%s (const %s *R /**< */)', field.c_iterator_name, c_type)
- _c('{')
- _c(' %s i;', field.c_iterator_type)
-
- if switch_obj is not None:
- _c(' i.data = %s;', fields[field.c_field_name][0])
- _c(' i.rem = %s;', _c_accessor_get_expr(field.type.expr, fields))
- elif field.prev_varsized_field == None:
- _c(' i.data = (%s *) (R + 1);', field.c_field_type)
- else:
- _c(' xcb_generic_iterator_t prev = %s;', _c_iterator_get_end(field.prev_varsized_field, 'R'))
- _c(' i.data = (%s *) ((char *) prev.data + XCB_TYPE_PAD(%s, prev.index));',
- field.c_field_type, field.c_field_type)
- if switch_obj is None:
- _c(' i.rem = %s;', _c_accessor_get_expr(field.type.expr, fields))
- _c(' i.index = (char *) i.data - (char *) %s;', 'R' if switch_obj is None else 'S' )
- _c(' return i;')
- _c('}')
-
-def _c_accessors(self, name, base):
- '''
- Declares the accessor functions for the fields of a structure.
- '''
- # no accessors for switch itself -
- # switch always needs to be unpacked explicitly
-# if self.is_switch:
-# pass
-# else:
- if True:
- for field in self.fields:
- if field.type.is_list and not field.type.fixed_size():
- _c_accessors_list(self, field)
- elif field.prev_varsized_field is not None or not field.type.fixed_size():
- _c_accessors_field(self, field)
-
-def c_simple(self, name):
- '''
- Exported function that handles cardinal type declarations.
- These are types which are typedef'd to one of the CARDx's, char, float, etc.
- '''
- _c_type_setup(self, name, ())
-
- if (self.name != name):
- # Typedef
- _h_setlevel(0)
- my_name = _t(name)
- _h('')
- _h('typedef %s %s;', _t(self.name), my_name)
-
- # Iterator
- _c_iterator(self, name)
-
-def _c_complex(self):
- '''
- Helper function for handling all structure types.
- Called for all structs, requests, replies, events, errors.
- '''
- _h_setlevel(0)
- _h('')
- _h('/**')
- _h(' * @brief %s', self.c_type)
- _h(' **/')
- _h('typedef %s %s {', self.c_container, self.c_type)
-
- struct_fields = []
- maxtypelen = 0
-
- varfield = None
- for field in self.fields:
- if not field.type.fixed_size() and not self.is_switch and not self.is_union:
- varfield = field.c_field_name
- continue
- if field.wire:
- struct_fields.append(field)
-
- for field in struct_fields:
- length = len(field.c_field_type)
- # account for '*' pointer_spec
- if not field.type.fixed_size():
- length += 1
- maxtypelen = max(maxtypelen, length)
-
- def _c_complex_field(self, field, space=''):
- if (field.type.fixed_size() or
- # in case of switch with switch children, don't make the field a pointer
- # necessary for unserialize to work
- (self.is_switch and field.type.is_switch)):
- spacing = ' ' * (maxtypelen - len(field.c_field_type))
- _h('%s %s%s %s%s; /**< */', space, field.c_field_type, spacing, field.c_field_name, field.c_subscript)
- else:
- spacing = ' ' * (maxtypelen - (len(field.c_field_type) + 1))
- _h('%s %s%s *%s%s; /**< */', space, field.c_field_type, spacing, field.c_field_name, field.c_subscript)
-
- if not self.is_switch:
- for field in struct_fields:
- _c_complex_field(self, field)
- else:
- for b in self.bitcases:
- space = ''
- if b.type.has_name:
- _h(' struct _%s {', b.c_field_name)
- space = ' '
- for field in b.type.fields:
- _c_complex_field(self, field, space)
- if b.type.has_name:
- _h(' } %s;', b.c_field_name)
-
- _h('} %s;', self.c_type)
-
-def c_struct(self, name):
- '''
- Exported function that handles structure declarations.
- '''
- _c_type_setup(self, name, ())
- _c_complex(self)
- _c_accessors(self, name, name)
- _c_iterator(self, name)
-
-def c_union(self, name):
- '''
- Exported function that handles union declarations.
- '''
- _c_type_setup(self, name, ())
- _c_complex(self)
- _c_iterator(self, name)
-
-def _c_request_helper(self, name, cookie_type, void, regular, aux=False):
- '''
- Declares a request function.
- '''
-
- # Four stunningly confusing possibilities here:
- #
- # Void Non-void
- # ------------------------------
- # "req" "req"
- # 0 flag CHECKED flag Normal Mode
- # void_cookie req_cookie
- # ------------------------------
- # "req_checked" "req_unchecked"
- # CHECKED flag 0 flag Abnormal Mode
- # void_cookie req_cookie
- # ------------------------------
-
-
- # Whether we are _checked or _unchecked
- checked = void and not regular
- unchecked = not void and not regular
-
- # What kind of cookie we return
- func_cookie = 'xcb_void_cookie_t' if void else self.c_cookie_type
-
- # What flag is passed to xcb_request
- func_flags = '0' if (void and regular) or (not void and not regular) else 'XCB_REQUEST_CHECKED'
-
- # Global extension id variable or NULL for xproto
- func_ext_global = '&' + _ns.c_ext_global_name if _ns.is_ext else '0'
-
- # What our function name is
- func_name = self.c_request_name if not aux else self.c_aux_name
- if checked:
- func_name = self.c_checked_name if not aux else self.c_aux_checked_name
- if unchecked:
- func_name = self.c_unchecked_name if not aux else self.c_aux_unchecked_name
-
- param_fields = []
- wire_fields = []
- maxtypelen = len('xcb_connection_t')
- serial_fields = []
- # special case: list with variable size elements
- list_with_var_size_elems = False
-
- for field in self.fields:
- if field.visible:
- # The field should appear as a call parameter
- param_fields.append(field)
- if field.wire and not field.auto:
- # We need to set the field up in the structure
- wire_fields.append(field)
- if field.type.need_serialize or field.type.need_sizeof:
- serial_fields.append(field)
-
- for field in param_fields:
- c_field_const_type = field.c_field_const_type
- if field.type.need_serialize and not aux:
- c_field_const_type = "const void"
- if len(c_field_const_type) > maxtypelen:
- maxtypelen = len(c_field_const_type)
- if field.type.is_list and not field.type.member.fixed_size():
- list_with_var_size_elems = True
-
- _h_setlevel(1)
- _c_setlevel(1)
- _h('')
- _h('/**')
- _h(' * Delivers a request to the X server')
- _h(' * @param c The connection')
- _h(' * @return A cookie')
- _h(' *')
- _h(' * Delivers a request to the X server.')
- _h(' * ')
- if checked:
- _h(' * This form can be used only if the request will not cause')
- _h(' * a reply to be generated. Any returned error will be')
- _h(' * saved for handling by xcb_request_check().')
- if unchecked:
- _h(' * This form can be used only if the request will cause')
- _h(' * a reply to be generated. Any returned error will be')
- _h(' * placed in the event queue.')
- _h(' */')
- _c('')
- _hc('')
- _hc('/*****************************************************************************')
- _hc(' **')
- _hc(' ** %s %s', cookie_type, func_name)
- _hc(' ** ')
-
- spacing = ' ' * (maxtypelen - len('xcb_connection_t'))
- _hc(' ** @param xcb_connection_t%s *c', spacing)
-
- for field in param_fields:
- c_field_const_type = field.c_field_const_type
- if field.type.need_serialize and not aux:
- c_field_const_type = "const void"
- spacing = ' ' * (maxtypelen - len(c_field_const_type))
- _hc(' ** @param %s%s %s%s', c_field_const_type, spacing, field.c_pointer, field.c_field_name)
-
- _hc(' ** @returns %s', cookie_type)
- _hc(' **')
- _hc(' *****************************************************************************/')
- _hc(' ')
- _hc('%s', cookie_type)
-
- spacing = ' ' * (maxtypelen - len('xcb_connection_t'))
- comma = ',' if len(param_fields) else ');'
- _h('%s (xcb_connection_t%s *c /**< */%s', func_name, spacing, comma)
- comma = ',' if len(param_fields) else ')'
- _c('%s (xcb_connection_t%s *c /**< */%s', func_name, spacing, comma)
-
- func_spacing = ' ' * (len(func_name) + 2)
- count = len(param_fields)
- for field in param_fields:
- count = count - 1
- c_field_const_type = field.c_field_const_type
- c_pointer = field.c_pointer
- if field.type.need_serialize and not aux:
- c_field_const_type = "const void"
- c_pointer = '*'
- spacing = ' ' * (maxtypelen - len(c_field_const_type))
- comma = ',' if count else ');'
- _h('%s%s%s %s%s /**< */%s', func_spacing, c_field_const_type,
- spacing, c_pointer, field.c_field_name, comma)
- comma = ',' if count else ')'
- _c('%s%s%s %s%s /**< */%s', func_spacing, c_field_const_type,
- spacing, c_pointer, field.c_field_name, comma)
-
- count = 2
- if not self.var_followed_by_fixed_fields:
- for field in param_fields:
- if not field.type.fixed_size():
- count = count + 2
- if field.type.need_serialize:
- # _serialize() keeps track of padding automatically
- count -= 1
- dimension = count + 2
-
- _c('{')
- _c(' static const xcb_protocol_request_t xcb_req = {')
- _c(' /* count */ %d,', count)
- _c(' /* ext */ %s,', func_ext_global)
- _c(' /* opcode */ %s,', self.c_request_name.upper())
- _c(' /* isvoid */ %d', 1 if void else 0)
- _c(' };')
- _c(' ')
-
- _c(' struct iovec xcb_parts[%d];', dimension)
- _c(' %s xcb_ret;', func_cookie)
- _c(' %s xcb_out;', self.c_type)
- if self.var_followed_by_fixed_fields:
- _c(' /* in the protocol description, variable size fields are followed by fixed size fields */')
- _c(' void *xcb_aux = 0;')
-
-
- for idx, f in enumerate(serial_fields):
- if aux:
- _c(' void *xcb_aux%d = 0;' % (idx))
- if list_with_var_size_elems:
- _c(' unsigned int i;')
- _c(' unsigned int xcb_tmp_len;')
- _c(' char *xcb_tmp;')
- _c(' ')
- # simple request call tracing
-# _c(' printf("in function %s\\n");' % func_name)
-
- # fixed size fields
- for field in wire_fields:
- if field.type.fixed_size():
- if field.type.is_expr:
- _c(' xcb_out.%s = %s;', field.c_field_name, _c_accessor_get_expr(field.type.expr, None))
- elif field.type.is_pad:
- if field.type.nmemb == 1:
- _c(' xcb_out.%s = 0;', field.c_field_name)
- else:
- _c(' memset(xcb_out.%s, 0, %d);', field.c_field_name, field.type.nmemb)
- else:
- if field.type.nmemb == 1:
- _c(' xcb_out.%s = %s;', field.c_field_name, field.c_field_name)
- else:
- _c(' memcpy(xcb_out.%s, %s, %d);', field.c_field_name, field.c_field_name, field.type.nmemb)
-
- def get_serialize_args(type_obj, c_field_name, aux_var, context='serialize'):
- serialize_args = get_serialize_params(context, type_obj,
- c_field_name,
- aux_var)[2]
- return reduce(lambda x,y: "%s, %s" % (x,y), [a[2] for a in serialize_args])
-
- # calls in order to free dyn. all. memory
- free_calls = []
-
- _c(' ')
- if not self.var_followed_by_fixed_fields:
- _c(' xcb_parts[2].iov_base = (char *) &xcb_out;')
- _c(' xcb_parts[2].iov_len = sizeof(xcb_out);')
- _c(' xcb_parts[3].iov_base = 0;')
- _c(' xcb_parts[3].iov_len = -xcb_parts[2].iov_len & 3;')
-
- count = 4
-
- for field in param_fields:
- if not field.type.fixed_size():
- _c(' /* %s %s */', field.type.c_type, field.c_field_name)
- # default: simple cast to char *
- if not field.type.need_serialize and not field.type.need_sizeof:
- _c(' xcb_parts[%d].iov_base = (char *) %s;', count, field.c_field_name)
- if field.type.is_list:
- if field.type.member.fixed_size():
- _c(' xcb_parts[%d].iov_len = %s * sizeof(%s);', count,
- _c_accessor_get_expr(field.type.expr, None),
- field.type.member.c_wiretype)
- else:
- list_length = _c_accessor_get_expr(field.type.expr, None)
-
- length = ''
- _c(" xcb_parts[%d].iov_len = 0;" % count)
- _c(" xcb_tmp = (char *)%s;", field.c_field_name)
- _c(" for(i=0; i<%s; i++) {" % list_length)
- _c(" xcb_tmp_len = %s(xcb_tmp);" %
- (field.type.c_sizeof_name))
- _c(" xcb_parts[%d].iov_len += xcb_tmp_len;" % count)
- _c(" xcb_tmp += xcb_tmp_len;")
- _c(" }")
- else:
- # not supposed to happen
- raise Exception("unhandled variable size field %s" % field.c_field_name)
- else:
- if not aux:
- _c(' xcb_parts[%d].iov_base = (char *) %s;', count, field.c_field_name)
- idx = serial_fields.index(field)
- aux_var = '&xcb_aux%d' % idx
- context = 'serialize' if aux else 'sizeof'
- _c(' xcb_parts[%d].iov_len = ', count)
- if aux:
- serialize_args = get_serialize_args(field.type, aux_var, field.c_field_name, context)
- _c(' %s (%s);', field.type.c_serialize_name, serialize_args)
- _c(' xcb_parts[%d].iov_base = xcb_aux%d;' % (count, idx))
- free_calls.append(' free(xcb_aux%d);' % idx)
- else:
- serialize_args = get_serialize_args(field.type, field.c_field_name, aux_var, context)
- func_name = field.type.c_sizeof_name
- _c(' %s (%s);', func_name, serialize_args)
-
- count += 1
- if not (field.type.need_serialize or field.type.need_sizeof):
- # the _serialize() function keeps track of padding automatically
- _c(' xcb_parts[%d].iov_base = 0;', count)
- _c(' xcb_parts[%d].iov_len = -xcb_parts[%d].iov_len & 3;', count, count-1)
- count += 1
-
- # elif self.var_followed_by_fixed_fields:
- else:
- _c(' xcb_parts[2].iov_base = (char *) &xcb_out;')
- # request header: opcodes + length
- _c(' xcb_parts[2].iov_len = 2*sizeof(uint8_t) + sizeof(uint16_t);')
- count += 1
- # call _serialize()
- buffer_var = '&xcb_aux'
- serialize_args = get_serialize_args(self, buffer_var, '&xcb_out', 'serialize')
- _c(' xcb_parts[%d].iov_len = %s (%s);', count, self.c_serialize_name, serialize_args)
- _c(' xcb_parts[%d].iov_base = (char *) xcb_aux;', count)
- free_calls.append(' free(xcb_aux);')
- # no padding necessary - _serialize() keeps track of padding automatically
-
- _c(' ')
- _c(' xcb_ret.sequence = xcb_send_request(c, %s, xcb_parts + 2, &xcb_req);', func_flags)
-
- # free dyn. all. data, if any
- for f in free_calls:
- _c(f)
- _c(' return xcb_ret;')
- _c('}')
-
-def _c_reply(self, name):
- '''
- Declares the function that returns the reply structure.
- '''
- spacing1 = ' ' * (len(self.c_cookie_type) - len('xcb_connection_t'))
- spacing2 = ' ' * (len(self.c_cookie_type) - len('xcb_generic_error_t'))
- spacing3 = ' ' * (len(self.c_reply_name) + 2)
-
- # check if _unserialize() has to be called for any field
- def look_for_special_cases(complex_obj):
- unserialize_fields = []
- # no unserialize call in case of switch
- if not complex_obj.is_switch:
- for field in complex_obj.fields:
- # three cases: 1. field with special case
- # 2. container that contains special case field
- # 3. list with special case elements
- if field.type.var_followed_by_fixed_fields:
- unserialize_fields.append(field)
- elif field.type.is_container:
- unserialize_fields += look_for_special_cases(field.type)
- elif field.type.is_list:
- if field.type.member.var_followed_by_fixed_fields:
- unserialize_fields.append(field)
- if field.type.member.is_container:
- unserialize_fields += look_for_special_cases(field.type.member)
- return unserialize_fields
-
- unserialize_fields = look_for_special_cases(self.reply)
-
- _h('')
- _h('/**')
- _h(' * Return the reply')
- _h(' * @param c The connection')
- _h(' * @param cookie The cookie')
- _h(' * @param e The xcb_generic_error_t supplied')
- _h(' *')
- _h(' * Returns the reply of the request asked by')
- _h(' * ')
- _h(' * The parameter @p e supplied to this function must be NULL if')
- _h(' * %s(). is used.', self.c_unchecked_name)
- _h(' * Otherwise, it stores the error if any.')
- _h(' *')
- _h(' * The returned value must be freed by the caller using free().')
- _h(' */')
- _c('')
- _hc('')
- _hc('/*****************************************************************************')
- _hc(' **')
- _hc(' ** %s * %s', self.c_reply_type, self.c_reply_name)
- _hc(' ** ')
- _hc(' ** @param xcb_connection_t%s *c', spacing1)
- _hc(' ** @param %s cookie', self.c_cookie_type)
- _hc(' ** @param xcb_generic_error_t%s **e', spacing2)
- _hc(' ** @returns %s *', self.c_reply_type)
- _hc(' **')
- _hc(' *****************************************************************************/')
- _hc(' ')
- _hc('%s *', self.c_reply_type)
- _hc('%s (xcb_connection_t%s *c /**< */,', self.c_reply_name, spacing1)
- _hc('%s%s cookie /**< */,', spacing3, self.c_cookie_type)
- _h('%sxcb_generic_error_t%s **e /**< */);', spacing3, spacing2)
- _c('%sxcb_generic_error_t%s **e /**< */)', spacing3, spacing2)
- _c('{')
-
- if len(unserialize_fields)>0:
- # certain variable size fields need to be unserialized explicitly
- _c(' %s *reply = (%s *) xcb_wait_for_reply(c, cookie.sequence, e);',
- self.c_reply_type, self.c_reply_type)
- _c(' int i;')
- for field in unserialize_fields:
- if field.type.is_list:
- _c(' %s %s_iter = %s(reply);', field.c_iterator_type, field.c_field_name, field.c_iterator_name)
- _c(' int %s_len = %s(reply);', field.c_field_name, field.c_length_name)
- _c(' %s *%s_data;', field.c_field_type, field.c_field_name)
- else:
- raise Exception('not implemented: call _unserialize() in reply for non-list type %s', field.c_field_type)
- # call _unserialize(), using the reply as source and target buffer
- _c(' /* special cases: transform parts of the reply to match XCB data structures */')
- for field in unserialize_fields:
- if field.type.is_list:
- _c(' for(i=0; i<%s_len; i++) {', field.c_field_name)
- _c(' %s_data = %s_iter.data;', field.c_field_name, field.c_field_name)
- _c(' %s((const void *)%s_data, &%s_data);', field.type.c_unserialize_name,
- field.c_field_name, field.c_field_name)
- _c(' %s(&%s_iter);', field.type.c_next_name, field.c_field_name)
- _c(' }')
- # return the transformed reply
- _c(' return reply;')
-
- else:
- _c(' return (%s *) xcb_wait_for_reply(c, cookie.sequence, e);', self.c_reply_type)
-
- _c('}')
-
-def _c_opcode(name, opcode):
- '''
- Declares the opcode define for requests, events, and errors.
- '''
- _h_setlevel(0)
- _h('')
- _h('/** Opcode for %s. */', _n(name))
- _h('#define %s %s', _n(name).upper(), opcode)
-
-def _c_cookie(self, name):
- '''
- Declares the cookie type for a non-void request.
- '''
- _h_setlevel(0)
- _h('')
- _h('/**')
- _h(' * @brief %s', self.c_cookie_type)
- _h(' **/')
- _h('typedef struct %s {', self.c_cookie_type)
- _h(' unsigned int sequence; /**< */')
- _h('} %s;', self.c_cookie_type)
-
-def c_request(self, name):
- '''
- Exported function that handles request declarations.
- '''
- _c_type_setup(self, name, ('request',))
-
- if self.reply:
- # Cookie type declaration
- _c_cookie(self, name)
-
- # Opcode define
- _c_opcode(name, self.opcode)
-
- # Request structure declaration
- _c_complex(self)
-
- if self.reply:
- _c_type_setup(self.reply, name, ('reply',))
- # Reply structure definition
- _c_complex(self.reply)
- # Request prototypes
- _c_request_helper(self, name, self.c_cookie_type, False, True)
- _c_request_helper(self, name, self.c_cookie_type, False, False)
- if self.need_aux:
- _c_request_helper(self, name, self.c_cookie_type, False, True, True)
- _c_request_helper(self, name, self.c_cookie_type, False, False, True)
- # Reply accessors
- _c_accessors(self.reply, name + ('reply',), name)
- _c_reply(self, name)
- else:
- # Request prototypes
- _c_request_helper(self, name, 'xcb_void_cookie_t', True, False)
- _c_request_helper(self, name, 'xcb_void_cookie_t', True, True)
- if self.need_aux:
- _c_request_helper(self, name, 'xcb_void_cookie_t', True, False, True)
- _c_request_helper(self, name, 'xcb_void_cookie_t', True, True, True)
-
-
-def c_event(self, name):
- '''
- Exported function that handles event declarations.
- '''
- _c_type_setup(self, name, ('event',))
-
- # Opcode define
- _c_opcode(name, self.opcodes[name])
-
- if self.name == name:
- # Structure definition
- _c_complex(self)
- else:
- # Typedef
- _h('')
- _h('typedef %s %s;', _t(self.name + ('event',)), _t(name + ('event',)))
-
-def c_error(self, name):
- '''
- Exported function that handles error declarations.
- '''
- _c_type_setup(self, name, ('error',))
-
- # Opcode define
- _c_opcode(name, self.opcodes[name])
-
- if self.name == name:
- # Structure definition
- _c_complex(self)
- else:
- # Typedef
- _h('')
- _h('typedef %s %s;', _t(self.name + ('error',)), _t(name + ('error',)))
-
-
-# Main routine starts here
-
-# Must create an "output" dictionary before any xcbgen imports.
-output = {'open' : c_open,
- 'close' : c_close,
- 'simple' : c_simple,
- 'enum' : c_enum,
- 'struct' : c_struct,
- 'union' : c_union,
- 'request' : c_request,
- 'event' : c_event,
- 'error' : c_error,
- }
-
-# Boilerplate below this point
-
-# Check for the argument that specifies path to the xcbgen python package.
-try:
- opts, args = getopt.getopt(sys.argv[1:], 'p:')
-except getopt.GetoptError, err:
- print str(err)
- print 'Usage: c_client.py [-p path] file.xml'
- sys.exit(1)
-
-for (opt, arg) in opts:
- if opt == '-p':
- sys.path.append(arg)
-
-# Import the module class
-try:
- from xcbgen.state import Module
-except ImportError:
- print ''
- print 'Failed to load the xcbgen Python package!'
- print 'Make sure that xcb/proto installed it on your Python path.'
- print 'If not, you will need to create a .pth file or define $PYTHONPATH'
- print 'to extend the path.'
- print 'Refer to the README file in xcb/proto for more info.'
- print ''
- raise
-
-# Parse the xml header
-module = Module(args[0], output)
-
-# Build type-registry and resolve type dependencies
-module.register()
-module.resolve()
-
-# Output the code
-module.generate()
+#!/usr/bin/env python +from xml.etree.cElementTree import * +from os.path import basename +import getopt +import sys +import re + +# Jump to the bottom of this file for the main routine + +# Some hacks to make the API more readable, and to keep backwards compability +_cname_re = re.compile('([A-Z0-9][a-z]+|[A-Z0-9]+(?![a-z])|[a-z]+)') +_cname_special_cases = {'DECnet':'decnet'} + +_extension_special_cases = ['XPrint', 'XCMisc', 'BigRequests'] + +_cplusplus_annoyances = {'class' : '_class', + 'new' : '_new', + 'delete': '_delete'} +_c_keywords = {'default' : '_default'} + +_hlines = [] +_hlevel = 0 +_clines = [] +_clevel = 0 +_ns = None + +# global variable to keep track of serializers and +# switch data types due to weird dependencies +finished_serializers = [] +finished_sizeof = [] +finished_switch = [] + +def _h(fmt, *args): + ''' + Writes the given line to the header file. + ''' + _hlines[_hlevel].append(fmt % args) + +def _c(fmt, *args): + ''' + Writes the given line to the source file. + ''' + _clines[_clevel].append(fmt % args) + +def _hc(fmt, *args): + ''' + Writes the given line to both the header and source files. + ''' + _h(fmt, *args) + _c(fmt, *args) + +# XXX See if this level thing is really necessary. +def _h_setlevel(idx): + ''' + Changes the array that header lines are written to. + Supports writing different sections of the header file. + ''' + global _hlevel + while len(_hlines) <= idx: + _hlines.append([]) + _hlevel = idx + +def _c_setlevel(idx): + ''' + Changes the array that source lines are written to. + Supports writing to different sections of the source file. + ''' + global _clevel + while len(_clines) <= idx: + _clines.append([]) + _clevel = idx + +def _n_item(str): + ''' + Does C-name conversion on a single string fragment. + Uses a regexp with some hard-coded special cases. + ''' + if str in _cname_special_cases: + return _cname_special_cases[str] + else: + split = _cname_re.finditer(str) + name_parts = [match.group(0) for match in split] + return '_'.join(name_parts) + +def _cpp(str): + ''' + Checks for certain C++ reserved words and fixes them. + ''' + if str in _cplusplus_annoyances: + return _cplusplus_annoyances[str] + elif str in _c_keywords: + return _c_keywords[str] + else: + return str + +def _ext(str): + ''' + Does C-name conversion on an extension name. + Has some additional special cases on top of _n_item. + ''' + if str in _extension_special_cases: + return _n_item(str).lower() + else: + return str.lower() + +def _n(list): + ''' + Does C-name conversion on a tuple of strings. + Different behavior depending on length of tuple, extension/not extension, etc. + Basically C-name converts the individual pieces, then joins with underscores. + ''' + if len(list) == 1: + parts = list + elif len(list) == 2: + parts = [list[0], _n_item(list[1])] + elif _ns.is_ext: + parts = [list[0], _ext(list[1])] + [_n_item(i) for i in list[2:]] + else: + parts = [list[0]] + [_n_item(i) for i in list[1:]] + return '_'.join(parts).lower() + +def _t(list): + ''' + Does C-name conversion on a tuple of strings representing a type. + Same as _n but adds a "_t" on the end. + ''' + if len(list) == 1: + parts = list + elif len(list) == 2: + parts = [list[0], _n_item(list[1]), 't'] + elif _ns.is_ext: + parts = [list[0], _ext(list[1])] + [_n_item(i) for i in list[2:]] + ['t'] + else: + parts = [list[0]] + [_n_item(i) for i in list[1:]] + ['t'] + return '_'.join(parts).lower() + + +def c_open(self): + ''' + Exported function that handles module open. + Opens the files and writes out the auto-generated comment, header file includes, etc. + ''' + global _ns + _ns = self.namespace + _ns.c_ext_global_name = _n(_ns.prefix + ('id',)) + + # Build the type-name collision avoidance table used by c_enum + build_collision_table() + + _h_setlevel(0) + _c_setlevel(0) + + _hc('/*') + _hc(' * This file generated automatically from %s by c_client.py.', _ns.file) + _hc(' * Edit at your peril.') + _hc(' */') + _hc('') + + _h('/**') + _h(' * @defgroup XCB_%s_API XCB %s API', _ns.ext_name, _ns.ext_name) + _h(' * @brief %s XCB Protocol Implementation.', _ns.ext_name) + _h(' * @{') + _h(' **/') + _h('') + _h('#ifndef __%s_H', _ns.header.upper()) + _h('#define __%s_H', _ns.header.upper()) + _h('') + _h('#include "xcb.h"') + + _c('#include <stdlib.h>') + _c('#include <string.h>') + _c('#include <assert.h>') + _c('#include "xcbext.h"') + _c('#include "%s.h"', _ns.header) + + if _ns.is_ext: + for (n, h) in self.imports: + _hc('#include "%s.h"', h) + + _h('') + _h('#ifdef __cplusplus') + _h('extern "C" {') + _h('#endif') + + if _ns.is_ext: + _h('') + _h('#define XCB_%s_MAJOR_VERSION %s', _ns.ext_name.upper(), _ns.major_version) + _h('#define XCB_%s_MINOR_VERSION %s', _ns.ext_name.upper(), _ns.minor_version) + _h(' ') #XXX + _h('extern xcb_extension_t %s;', _ns.c_ext_global_name) + + _c('') + _c('xcb_extension_t %s = { "%s", 0 };', _ns.c_ext_global_name, _ns.ext_xname) + +def c_close(self): + ''' + Exported function that handles module close. + Writes out all the stored content lines, then closes the files. + ''' + _h_setlevel(2) + _c_setlevel(2) + _hc('') + + _h('') + _h('#ifdef __cplusplus') + _h('}') + _h('#endif') + + _h('') + _h('#endif') + _h('') + _h('/**') + _h(' * @}') + _h(' */') + + # Write header file + hfile = open('%s.h' % _ns.header, 'w') + for list in _hlines: + for line in list: + hfile.write(line) + hfile.write('\n') + hfile.close() + + # Write source file + cfile = open('%s.c' % _ns.header, 'w') + for list in _clines: + for line in list: + cfile.write(line) + cfile.write('\n') + cfile.close() + +def build_collision_table(): + global namecount + namecount = {} + + for v in module.types.values(): + name = _t(v[0]) + namecount[name] = (namecount.get(name) or 0) + 1 + +def c_enum(self, name): + ''' + Exported function that handles enum declarations. + ''' + + tname = _t(name) + if namecount[tname] > 1: + tname = _t(name + ('enum',)) + + _h_setlevel(0) + _h('') + _h('typedef enum %s {', tname) + + count = len(self.values) + + for (enam, eval) in self.values: + count = count - 1 + equals = ' = ' if eval != '' else '' + comma = ',' if count > 0 else '' + _h(' %s%s%s%s', _n(name + (enam,)).upper(), equals, eval, comma) + + _h('} %s;', tname) + +def _c_type_setup(self, name, postfix): + ''' + Sets up all the C-related state by adding additional data fields to + all Field and Type objects. Here is where we figure out most of our + variable and function names. + + Recurses into child fields and list member types. + ''' + # Do all the various names in advance + self.c_type = _t(name + postfix) + self.c_wiretype = 'char' if self.c_type == 'void' else self.c_type + + self.c_iterator_type = _t(name + ('iterator',)) + self.c_next_name = _n(name + ('next',)) + self.c_end_name = _n(name + ('end',)) + + self.c_request_name = _n(name) + self.c_checked_name = _n(name + ('checked',)) + self.c_unchecked_name = _n(name + ('unchecked',)) + self.c_reply_name = _n(name + ('reply',)) + self.c_reply_type = _t(name + ('reply',)) + self.c_cookie_type = _t(name + ('cookie',)) + + self.need_aux = False + self.need_serialize = False + self.need_sizeof = False + + self.c_aux_name = _n(name + ('aux',)) + self.c_aux_checked_name = _n(name + ('aux', 'checked')) + self.c_aux_unchecked_name = _n(name + ('aux', 'unchecked')) + self.c_serialize_name = _n(name + ('serialize',)) + self.c_unserialize_name = _n(name + ('unserialize',)) + self.c_unpack_name = _n(name + ('unpack',)) + self.c_sizeof_name = _n(name + ('sizeof',)) + + # special case: structs where variable size fields are followed by fixed size fields + self.var_followed_by_fixed_fields = False + + if self.is_switch: + self.need_serialize = True + self.c_container = 'struct' + for bitcase in self.bitcases: + bitcase.c_field_name = _cpp(bitcase.field_name) + bitcase_name = bitcase.field_type if bitcase.type.has_name else name + _c_type_setup(bitcase.type, bitcase_name, ()) + + elif self.is_container: + + self.c_container = 'union' if self.is_union else 'struct' + prev_varsized_field = None + prev_varsized_offset = 0 + first_field_after_varsized = None + + for field in self.fields: + _c_type_setup(field.type, field.field_type, ()) + if field.type.is_list: + _c_type_setup(field.type.member, field.field_type, ()) + if (field.type.nmemb is None): + self.need_sizeof = True + + field.c_field_type = _t(field.field_type) + field.c_field_const_type = ('' if field.type.nmemb == 1 else 'const ') + field.c_field_type + field.c_field_name = _cpp(field.field_name) + field.c_subscript = '[%d]' % field.type.nmemb if (field.type.nmemb > 1) else '' + field.c_pointer = ' ' if field.type.nmemb == 1 else '*' + + # correct the c_pointer field for variable size non-list types + if not field.type.fixed_size() and field.c_pointer == ' ': + field.c_pointer = '*' + if field.type.is_list and not field.type.member.fixed_size(): + field.c_pointer = '*' + + if field.type.is_switch: + field.c_pointer = '*' + field.c_field_const_type = 'const ' + field.c_field_type + self.need_aux = True + elif not field.type.fixed_size() and not field.type.is_bitcase: + self.need_sizeof = True + + field.c_iterator_type = _t(field.field_type + ('iterator',)) # xcb_fieldtype_iterator_t + field.c_iterator_name = _n(name + (field.field_name, 'iterator')) # xcb_container_field_iterator + field.c_accessor_name = _n(name + (field.field_name,)) # xcb_container_field + field.c_length_name = _n(name + (field.field_name, 'length')) # xcb_container_field_length + field.c_end_name = _n(name + (field.field_name, 'end')) # xcb_container_field_end + + field.prev_varsized_field = prev_varsized_field + field.prev_varsized_offset = prev_varsized_offset + + if prev_varsized_offset == 0: + first_field_after_varsized = field + field.first_field_after_varsized = first_field_after_varsized + + if field.type.fixed_size(): + prev_varsized_offset += field.type.size + # special case: intermixed fixed and variable size fields + if prev_varsized_field is not None and not field.type.is_pad and field.wire: + if not self.is_union: + self.need_serialize = True + self.var_followed_by_fixed_fields = True + else: + self.last_varsized_field = field + prev_varsized_field = field + prev_varsized_offset = 0 + + if self.var_followed_by_fixed_fields: + if field.type.fixed_size(): + field.prev_varsized_field = None + + if self.need_serialize: + # when _unserialize() is wanted, create _sizeof() as well for consistency reasons + self.need_sizeof = True + + # as switch does never appear at toplevel, + # continue here with type construction + if self.is_switch: + if self.c_type not in finished_switch: + finished_switch.append(self.c_type) + # special: switch C structs get pointer fields for variable-sized members + _c_complex(self) + for bitcase in self.bitcases: + bitcase_name = bitcase.type.name if bitcase.type.has_name else name + _c_accessors(bitcase.type, bitcase_name, bitcase_name) + # no list with switch as element, so no call to + # _c_iterator(field.type, field_name) necessary + + if not self.is_bitcase: + if self.need_serialize: + if self.c_serialize_name not in finished_serializers: + finished_serializers.append(self.c_serialize_name) + _c_serialize('serialize', self) + + # _unpack() and _unserialize() are only needed for special cases: + # switch -> unpack + # special cases -> unserialize + if self.is_switch or self.var_followed_by_fixed_fields: + _c_serialize('unserialize', self) + + if self.need_sizeof: + if self.c_sizeof_name not in finished_sizeof: + if not module.namespace.is_ext or self.name[:2] == module.namespace.prefix: + finished_sizeof.append(self.c_sizeof_name) + _c_serialize('sizeof', self) +# _c_type_setup() + +def _c_helper_absolute_name(prefix, field=None): + """ + turn prefix, which is a list of tuples (name, separator, Type obj) into a string + representing a valid name in C (based on the context) + if field is not None, append the field name as well + """ + prefix_str = '' + for name, sep, obj in prefix: + prefix_str += name + if '' == sep: + sep = '->' + if ((obj.is_bitcase and obj.has_name) or # named bitcase + (obj.is_switch and len(obj.parents)>1)): + sep = '.' + prefix_str += sep + if field is not None: + prefix_str += _cpp(field.field_name) + return prefix_str +# _c_absolute_name + +def _c_helper_field_mapping(complex_type, prefix, flat=False): + """ + generate absolute names, based on prefix, for all fields starting from complex_type + if flat == True, nested complex types are not taken into account + """ + all_fields = {} + if complex_type.is_switch: + for b in complex_type.bitcases: + if b.type.has_name: + switch_name, switch_sep, switch_type = prefix[-1] + bitcase_prefix = prefix + [(b.type.name[-1], '.', b.type)] + else: + bitcase_prefix = prefix + + if (True==flat and not b.type.has_name) or False==flat: + all_fields.update(_c_helper_field_mapping(b.type, bitcase_prefix, flat)) + else: + for f in complex_type.fields: + fname = _c_helper_absolute_name(prefix, f) + if all_fields.has_key(f.field_name): + raise Exception("field name %s has been registered before" % f.field_name) + + all_fields[f.field_name] = (fname, f) + if f.type.is_container and flat==False: + if f.type.is_bitcase and not f.type.has_name: + new_prefix = prefix + elif f.type.is_switch and len(f.type.parents)>1: + # nested switch gets another separator + new_prefix = prefix+[(f.c_field_name, '.', f.type)] + else: + new_prefix = prefix+[(f.c_field_name, '->', f.type)] + all_fields.update(_c_helper_field_mapping(f.type, new_prefix, flat)) + + return all_fields +# _c_field_mapping() + +def _c_helper_resolve_field_names (prefix): + """ + get field names for all objects in the prefix array + """ + all_fields = {} + tmp_prefix = [] + # look for fields in the remaining containers + for idx, p in enumerate(prefix): + name, sep, obj = p + if ''==sep: + # sep can be preset in prefix, if not, make a sensible guess + sep = '.' if (obj.is_switch or obj.is_bitcase) else '->' + # exception: 'toplevel' object (switch as well!) always have sep '->' + sep = '->' if idx<1 else sep + if not obj.is_bitcase or (obj.is_bitcase and obj.has_name): + tmp_prefix.append((name, sep, obj)) + all_fields.update(_c_helper_field_mapping(obj, tmp_prefix, flat=True)) + + return all_fields +# _c_helper_resolve_field_names + +def get_expr_fields(self): + """ + get the Fields referenced by switch or list expression + """ + def get_expr_field_names(expr): + if expr.op is None: + if expr.lenfield_name is not None: + return [expr.lenfield_name] + else: + # constant value expr + return [] + else: + if expr.op == '~': + return get_expr_field_names(expr.rhs) + elif expr.op == 'popcount': + return get_expr_field_names(expr.rhs) + elif expr.op == 'sumof': + # sumof expr references another list, + # we need that list's length field here + field = None + for f in expr.lenfield_parent.fields: + if f.field_name == expr.lenfield_name: + field = f + break + if field is None: + raise Exception("list field '%s' referenced by sumof not found" % expr.lenfield_name) + # referenced list + its length field + return [expr.lenfield_name] + get_expr_field_names(field.type.expr) + elif expr.op == 'enumref': + return [] + else: + return get_expr_field_names(expr.lhs) + get_expr_field_names(expr.rhs) + # get_expr_field_names() + + # resolve the field names with the parent structure(s) + unresolved_fields_names = get_expr_field_names(self.expr) + + # construct prefix from self + prefix = [('', '', p) for p in self.parents] + if self.is_container: + prefix.append(('', '', self)) + + all_fields = _c_helper_resolve_field_names (prefix) + resolved_fields_names = filter(lambda x: x in all_fields.keys(), unresolved_fields_names) + if len(unresolved_fields_names) != len(resolved_fields_names): + raise Exception("could not resolve all fields for %s" % self.name) + + resolved_fields = [all_fields[n][1] for n in resolved_fields_names] + return resolved_fields +# get_expr_fields() + +def resolve_expr_fields(complex_obj): + """ + find expr fields appearing in complex_obj and descendents that cannot be resolved within complex_obj + these are normally fields that need to be given as function parameters + """ + all_fields = [] + expr_fields = [] + unresolved = [] + + for field in complex_obj.fields: + all_fields.append(field) + if field.type.is_switch or field.type.is_list: + expr_fields += get_expr_fields(field.type) + if field.type.is_container: + expr_fields += resolve_expr_fields(field.type) + + # try to resolve expr fields + for e in expr_fields: + if e not in all_fields and e not in unresolved: + unresolved.append(e) + return unresolved +# resolve_expr_fields() + +def get_serialize_params(context, self, buffer_var='_buffer', aux_var='_aux'): + """ + functions like _serialize(), _unserialize(), and _unpack() sometimes need additional parameters: + E.g. in order to unpack switch, extra parameters might be needed to evaluate the switch + expression. This function tries to resolve all fields within a structure, and returns the + unresolved fields as the list of external parameters. + """ + def add_param(params, param): + if param not in params: + params.append(param) + + # collect all fields into param_fields + param_fields = [] + wire_fields = [] + + for field in self.fields: + if field.visible: + # the field should appear as a parameter in the function call + param_fields.append(field) + if field.wire and not field.auto: + if field.type.fixed_size() and not self.is_switch: + # field in the xcb_out structure + wire_fields.append(field) + # fields like 'pad0' are skipped! + + # in case of switch, parameters always contain any fields referenced in the switch expr + # we do not need any variable size fields here, as the switch data type contains both + # fixed and variable size fields + if self.is_switch: + param_fields = get_expr_fields(self) + + # _serialize()/_unserialize()/_unpack() function parameters + # note: don't use set() for params, it is unsorted + params = [] + + # 1. the parameter for the void * buffer + if 'serialize' == context: + params.append(('void', '**', buffer_var)) + elif context in ('unserialize', 'unpack', 'sizeof'): + params.append(('const void', '*', buffer_var)) + + # 2. any expr fields that cannot be resolved within self and descendants + unresolved_fields = resolve_expr_fields(self) + for f in unresolved_fields: + add_param(params, (f.c_field_type, '', f.c_field_name)) + + # 3. param_fields contain the fields necessary to evaluate the switch expr or any other fields + # that do not appear in the data type struct + for p in param_fields: + if self.is_switch: + typespec = p.c_field_const_type + pointerspec = p.c_pointer + add_param(params, (typespec, pointerspec, p.c_field_name)) + else: + if p.visible and not p.wire and not p.auto: + typespec = p.c_field_type + pointerspec = '' + add_param(params, (typespec, pointerspec, p.c_field_name)) + + # 4. aux argument + if 'serialize' == context: + add_param(params, ('const %s' % self.c_type, '*', aux_var)) + elif 'unserialize' == context: + add_param(params, ('%s' % self.c_type, '**', aux_var)) + elif 'unpack' == context: + add_param(params, ('%s' % self.c_type, '*', aux_var)) + + # 5. switch contains all variable size fields as struct members + # for other data types though, these have to be supplied separately + # this is important for the special case of intermixed fixed and + # variable size fields + if not self.is_switch and 'serialize' == context: + for p in param_fields: + if not p.type.fixed_size(): + add_param(params, (p.c_field_const_type, '*', p.c_field_name)) + + return (param_fields, wire_fields, params) +# get_serialize_params() + +def _c_serialize_helper_insert_padding(context, code_lines, space, postpone): + code_lines.append('%s /* insert padding */' % space) + code_lines.append('%s xcb_pad = -xcb_block_len & 3;' % space) +# code_lines.append('%s printf("automatically inserting padding: %%%%d\\n", xcb_pad);' % space) + code_lines.append('%s xcb_buffer_len += xcb_block_len + xcb_pad;' % space) + + if not postpone: + code_lines.append('%s if (0 != xcb_pad) {' % space) + + if 'serialize' == context: + code_lines.append('%s xcb_parts[xcb_parts_idx].iov_base = xcb_pad0;' % space) + code_lines.append('%s xcb_parts[xcb_parts_idx].iov_len = xcb_pad;' % space) + code_lines.append('%s xcb_parts_idx++;' % space) + elif context in ('unserialize', 'unpack', 'sizeof'): + code_lines.append('%s xcb_tmp += xcb_pad;' % space) + + code_lines.append('%s xcb_pad = 0;' % space) + code_lines.append('%s }' % space) + + code_lines.append('%s xcb_block_len = 0;' % space) + + # keep tracking of xcb_parts entries for serialize + return 1 +# _c_serialize_helper_insert_padding() + +def _c_serialize_helper_switch(context, self, complex_name, + code_lines, temp_vars, + space, prefix): + count = 0 + switch_expr = _c_accessor_get_expr(self.expr, None) + + for b in self.bitcases: + bitcase_expr = _c_accessor_get_expr(b.type.expr, None) + code_lines.append(' if(%s & %s) {' % (switch_expr, bitcase_expr)) +# code_lines.append(' printf("switch %s: entering bitcase section %s (mask=%%%%d)...\\n", %s);' % +# (self.name[-1], b.type.name[-1], bitcase_expr)) + b_prefix = prefix + if b.type.has_name: + b_prefix = prefix + [(b.c_field_name, '.', b.type)] + + count += _c_serialize_helper_fields(context, b.type, + code_lines, temp_vars, + "%s " % space, + b_prefix, + is_bitcase = True) + code_lines.append(' }') + +# if 'serialize' == context: +# count += _c_serialize_helper_insert_padding(context, code_lines, space, False) +# elif context in ('unserialize', 'unpack', 'sizeof'): +# # padding +# code_lines.append('%s xcb_pad = -xcb_block_len & 3;' % space) +# code_lines.append('%s xcb_buffer_len += xcb_block_len + xcb_pad;' % space) + + return count +# _c_serialize_helper_switch + +def _c_serialize_helper_switch_field(context, self, field, c_switch_variable, prefix): + """ + handle switch by calling _serialize() or _unpack(), depending on context + """ + # switch is handled by this function as a special case + param_fields, wire_fields, params = get_serialize_params(context, self) + field_mapping = _c_helper_field_mapping(self, prefix) + prefix_str = _c_helper_absolute_name(prefix) + + # find the parameters that need to be passed to _serialize()/_unpack(): + # all switch expr fields must be given as parameters + args = get_expr_fields(field.type) + # length fields for variable size types in switch, normally only some of need + # need to be passed as parameters + switch_len_fields = resolve_expr_fields(field.type) + + # a switch field at this point _must_ be a bitcase field + # we require that bitcases are "self-contiguous" + bitcase_unresolved = resolve_expr_fields(self) + if len(bitcase_unresolved) != 0: + raise Exception('unresolved fields within bitcase is not supported at this point') + + # get the C names for the parameters + c_field_names = '' + for a in switch_len_fields: + c_field_names += "%s, " % field_mapping[a.c_field_name][0] + for a in args: + c_field_names += "%s, " % field_mapping[a.c_field_name][0] + + # call _serialize()/_unpack() to determine the actual size + if 'serialize' == context: + length = "%s(&%s, %s&%s%s)" % (field.type.c_serialize_name, c_switch_variable, + c_field_names, prefix_str, field.c_field_name) + elif context in ('unserialize', 'unpack'): + length = "%s(xcb_tmp, %s&%s%s)" % (field.type.c_unpack_name, + c_field_names, prefix_str, field.c_field_name) + + return length +# _c_serialize_helper_switch_field() + +def _c_serialize_helper_list_field(context, self, field, + code_lines, temp_vars, + space, prefix): + """ + helper function to cope with lists of variable length + """ + expr = field.type.expr + prefix_str = _c_helper_absolute_name(prefix) + param_fields, wire_fields, params = get_serialize_params('sizeof', self) + param_names = [p[2] for p in params] + + expr_fields_names = [f.field_name for f in get_expr_fields(field.type)] + resolved = filter(lambda x: x in param_names, expr_fields_names) + unresolved = filter(lambda x: x not in param_names, expr_fields_names) + + field_mapping = {} + for r in resolved: + field_mapping[r] = (r, None) + + if len(unresolved)>0: + tmp_prefix = prefix + if len(tmp_prefix)==0: + raise Exception("found an empty prefix while resolving expr field names for list %s", + field.c_field_name) + + field_mapping.update(_c_helper_resolve_field_names(prefix)) + resolved += filter(lambda x: x in field_mapping, unresolved) + unresolved = filter(lambda x: x not in field_mapping, unresolved) + if len(unresolved)>0: + raise Exception('could not resolve the length fields required for list %s' % field.c_field_name) + + list_length = _c_accessor_get_expr(expr, field_mapping) + + # default: list with fixed size elements + length = '%s * sizeof(%s)' % (list_length, field.type.member.c_wiretype) + + # list with variable-sized elements + if not field.type.member.fixed_size(): + length = '' + if context in ('unserialize', 'sizeof', 'unpack'): + int_i = ' unsigned int i;' + xcb_tmp_len = ' unsigned int xcb_tmp_len;' + if int_i not in temp_vars: + temp_vars.append(int_i) + if xcb_tmp_len not in temp_vars: + temp_vars.append(xcb_tmp_len) + # loop over all list elements and call sizeof repeatedly + # this should be a bit faster than using the iterators + code_lines.append("%s for(i=0; i<%s; i++) {" % (space, list_length)) + code_lines.append("%s xcb_tmp_len = %s(xcb_tmp);" % + (space, field.type.c_sizeof_name)) + code_lines.append("%s xcb_block_len += xcb_tmp_len;" % space) + code_lines.append("%s xcb_tmp += xcb_tmp_len;" % space) + code_lines.append("%s }" % space) + + elif 'serialize' == context: + code_lines.append('%s xcb_parts[xcb_parts_idx].iov_len = 0;' % space) + code_lines.append('%s xcb_tmp = (char *) %s%s;' % (space, prefix_str, field.c_field_name)) + code_lines.append('%s for(i=0; i<%s; i++) { ' % (space, list_length)) + code_lines.append('%s xcb_block_len = %s(xcb_tmp);' % (space, field.type.c_sizeof_name)) + code_lines.append('%s xcb_parts[xcb_parts_idx].iov_len += xcb_block_len;' % space) + code_lines.append('%s }' % space) + code_lines.append('%s xcb_block_len = xcb_parts[xcb_parts_idx].iov_len;' % space) + + return length +# _c_serialize_helper_list_field() + +def _c_serialize_helper_fields_fixed_size(context, self, field, + code_lines, temp_vars, + space, prefix): + # keep the C code a bit more readable by giving the field name + if not self.is_bitcase: + code_lines.append('%s /* %s.%s */' % (space, self.c_type, field.c_field_name)) + else: + scoped_name = [p[2].c_type if idx==0 else p[0] for idx, p in enumerate(prefix)] + typename = reduce(lambda x,y: "%s.%s" % (x, y), scoped_name) + code_lines.append('%s /* %s.%s */' % (space, typename, field.c_field_name)) + + abs_field_name = _c_helper_absolute_name(prefix, field) + # default for simple cases: call sizeof() + length = "sizeof(%s)" % field.c_field_type + + if context in ('unserialize', 'unpack', 'sizeof'): + # default: simple cast + value = ' %s = *(%s *)xcb_tmp;' % (abs_field_name, field.c_field_type) + + # padding - we could probably just ignore it + if field.type.is_pad and field.type.nmemb > 1: + value = '' + for i in range(field.type.nmemb): + code_lines.append('%s %s[%d] = *(%s *)xcb_tmp;' % + (space, abs_field_name, i, field.c_field_type)) + # total padding = sizeof(pad0) * nmemb + length += " * %d" % field.type.nmemb + + if field.type.is_list: + # no such case in the protocol, cannot be tested and therefore ignored for now + raise Exception('list with fixed number of elemens unhandled in _unserialize()') + + elif 'serialize' == context: + value = ' xcb_parts[xcb_parts_idx].iov_base = (char *) ' + + if field.type.is_expr: + # need to register a temporary variable for the expression in case we know its type + if field.type.c_type is None: + raise Exception("type for field '%s' (expression '%s') unkown" % + (field.field_name, _c_accessor_get_expr(field.type.expr))) + + temp_vars.append(' %s xcb_expr_%s = %s;' % (field.type.c_type, _cpp(field.field_name), + _c_accessor_get_expr(field.type.expr, prefix))) + value += "&xcb_expr_%s;" % _cpp(field.field_name) + + elif field.type.is_pad: + if field.type.nmemb == 1: + value += "&xcb_pad;" + else: + # we could also set it to 0, see definition of xcb_send_request() + value = ' xcb_parts[xcb_parts_idx].iov_base = xcb_pad0;' + length += "*%d" % field.type.nmemb + + else: + # non-list type with fixed size + if field.type.nmemb == 1: + value += "&%s;" % (abs_field_name) + + # list with nmemb (fixed size) elements + else: + value += '%s;' % (abs_field_name) + length = '%d' % field.type.nmemb + + return (value, length) +# _c_serialize_helper_fields_fixed_size() + +def _c_serialize_helper_fields_variable_size(context, self, field, + code_lines, temp_vars, + space, prefix): + prefix_str = _c_helper_absolute_name(prefix) + + if context in ('unserialize', 'unpack', 'sizeof'): + value = '' + var_field_name = 'xcb_tmp' + + # special case: intermixed fixed and variable size fields + if self.var_followed_by_fixed_fields and 'unserialize' == context: + value = ' %s = (%s *)xcb_tmp;' % (field.c_field_name, field.c_field_type) + temp_vars.append(' %s *%s;' % (field.type.c_type, field.c_field_name)) + # special case: switch + if 'unpack' == context: + value = ' %s%s = (%s *)xcb_tmp;' % (prefix_str, field.c_field_name, field.c_field_type) + + elif 'serialize' == context: + # variable size fields appear as parameters to _serialize() if the + # 'toplevel' container is not a switch + prefix_string = prefix_str if prefix[0][2].is_switch else '' + var_field_name = "%s%s" % (prefix_string, field.c_field_name) + value = ' xcb_parts[xcb_parts_idx].iov_base = (char *) %s;' % var_field_name + + length = '' + + code_lines.append('%s /* %s */' % (space, field.c_field_name)) + + if field.type.is_list: + if value != '': + # in any context, list is already a pointer, so the default assignment is ok + code_lines.append("%s%s" % (space, value)) + value = '' + length = _c_serialize_helper_list_field(context, self, field, + code_lines, temp_vars, + space, prefix) + + elif field.type.is_switch: + value = '' + if context == 'serialize': + # the _serialize() function allocates the correct amount memory if given a NULL pointer + value = ' xcb_parts[xcb_parts_idx].iov_base = (char *)0;' + length = _c_serialize_helper_switch_field(context, self, field, + 'xcb_parts[xcb_parts_idx].iov_base', + prefix) + + else: + # in all remaining special cases - call _sizeof() + length = "%s(%s)" % (field.type.c_sizeof_name, var_field_name) + + return (value, length) +# _c_serialize_helper_fields_variable_size + +def _c_serialize_helper_fields(context, self, + code_lines, temp_vars, + space, prefix, is_bitcase): + count = 0 + need_padding = False + prev_field_was_variable = False + + for field in self.fields: + if not field.visible: + if not ((field.wire and not field.auto) or 'unserialize' == context): + continue + + # switch/bitcase: fixed size fields must be considered explicitly + if field.type.fixed_size(): + if self.is_bitcase or self.var_followed_by_fixed_fields: + if prev_field_was_variable and need_padding: + # insert padding +# count += _c_serialize_helper_insert_padding(context, code_lines, space, +# self.var_followed_by_fixed_fields) + prev_field_was_variable = False + + # prefix for fixed size fields + fixed_prefix = prefix + + value, length = _c_serialize_helper_fields_fixed_size(context, self, field, + code_lines, temp_vars, + space, fixed_prefix) + else: + continue + + # fields with variable size + else: + # switch/bitcase: always calculate padding before and after variable sized fields + if need_padding or is_bitcase: + count += _c_serialize_helper_insert_padding(context, code_lines, space, + self.var_followed_by_fixed_fields) + + value, length = _c_serialize_helper_fields_variable_size(context, self, field, + code_lines, temp_vars, + space, prefix) + prev_field_was_variable = True + + # save (un)serialization C code + if '' != value: + code_lines.append('%s%s' % (space, value)) + + if field.type.fixed_size(): + if is_bitcase or self.var_followed_by_fixed_fields: + # keep track of (un)serialized object's size + code_lines.append('%s xcb_block_len += %s;' % (space, length)) + if context in ('unserialize', 'unpack', 'sizeof'): + code_lines.append('%s xcb_tmp += %s;' % (space, length)) + else: + # variable size objects or bitcase: + # value & length might have been inserted earlier for special cases + if '' != length: + # special case: intermixed fixed and variable size fields + if (not field.type.fixed_size() and + self.var_followed_by_fixed_fields and 'unserialize' == context): + temp_vars.append(' int %s_len;' % field.c_field_name) + code_lines.append('%s %s_len = %s;' % (space, field.c_field_name, length)) + code_lines.append('%s xcb_block_len += %s_len;' % (space, field.c_field_name)) + code_lines.append('%s xcb_tmp += %s_len;' % (space, field.c_field_name)) + else: + code_lines.append('%s xcb_block_len += %s;' % (space, length)) + # increase pointer into the byte stream accordingly + if context in ('unserialize', 'sizeof', 'unpack'): + code_lines.append('%s xcb_tmp += xcb_block_len;' % space) + + if 'serialize' == context: + if '' != length: + code_lines.append('%s xcb_parts[xcb_parts_idx].iov_len = %s;' % (space, length)) + code_lines.append('%s xcb_parts_idx++;' % space) + count += 1 + + need_padding = True + if self.var_followed_by_fixed_fields: + need_padding = False + + return count +# _c_serialize_helper_fields() + +def _c_serialize_helper(context, complex_type, + code_lines, temp_vars, + space='', prefix=[]): + # count tracks the number of fields to serialize + count = 0 + + if hasattr(complex_type, 'type'): + self = complex_type.type + complex_name = complex_type.name + else: + self = complex_type + if self.var_followed_by_fixed_fields and 'unserialize' == context: + complex_name = 'xcb_out' + else: + complex_name = '_aux' + + # special case: switch is serialized by evaluating each bitcase separately + if self.is_switch: + count += _c_serialize_helper_switch(context, self, complex_name, + code_lines, temp_vars, + space, prefix) + + # all other data types can be evaluated one field a time + else: + # unserialize & fixed size fields: simply cast the buffer to the respective xcb_out type + if context in ('unserialize', 'unpack', 'sizeof') and not self.var_followed_by_fixed_fields: + code_lines.append('%s xcb_block_len += sizeof(%s);' % (space, self.c_type)) + code_lines.append('%s xcb_tmp += xcb_block_len;' % space) + # probably not needed + #_c_serialize_helper_insert_padding(context, code_lines, space, False) + + count += _c_serialize_helper_fields(context, self, + code_lines, temp_vars, + space, prefix, False) + # "final padding" + count += _c_serialize_helper_insert_padding(context, code_lines, space, False) + + return count +# _c_serialize_helper() + +def _c_serialize(context, self): + """ + depending on the context variable, generate _serialize(), _unserialize(), _unpack(), or _sizeof() + for the ComplexType variable self + """ + _h_setlevel(1) + _c_setlevel(1) + + _hc('') + # _serialize() returns the buffer size + _hc('int') + + if self.is_switch and 'unserialize' == context: + context = 'unpack' + + cases = { 'serialize' : self.c_serialize_name, + 'unserialize' : self.c_unserialize_name, + 'unpack' : self.c_unpack_name, + 'sizeof' : self.c_sizeof_name } + func_name = cases[context] + + param_fields, wire_fields, params = get_serialize_params(context, self) + variable_size_fields = 0 + # maximum space required for type definition of function arguments + maxtypelen = 0 + + # determine N(variable_fields) + for field in param_fields: + # if self.is_switch, treat all fields as if they are variable sized + if not field.type.fixed_size() or self.is_switch: + variable_size_fields += 1 + # determine maxtypelen + for p in params: + maxtypelen = max(maxtypelen, len(p[0]) + len(p[1])) + + # write to .c/.h + indent = ' '*(len(func_name)+2) + param_str = [] + for p in params: + typespec, pointerspec, field_name = p + spacing = ' '*(maxtypelen-len(typespec)-len(pointerspec)) + param_str.append("%s%s%s %s%s /**< */" % (indent, typespec, spacing, pointerspec, field_name)) + # insert function name + param_str[0] = "%s (%s" % (func_name, param_str[0].strip()) + param_str = map(lambda x: "%s," % x, param_str) + for s in param_str[:-1]: + _hc(s) + _h("%s);" % param_str[-1].rstrip(',')) + _c("%s)" % param_str[-1].rstrip(',')) + _c('{') + + code_lines = [] + temp_vars = [] + prefix = [] + + if 'serialize' == context: + if not self.is_switch and not self.var_followed_by_fixed_fields: + _c(' %s *xcb_out = *_buffer;', self.c_type) + _c(' unsigned int xcb_out_pad = -sizeof(%s) & 3;', self.c_type) + _c(' unsigned int xcb_buffer_len = sizeof(%s) + xcb_out_pad;', self.c_type) + else: + _c(' char *xcb_out = *_buffer;') + _c(' unsigned int xcb_buffer_len = 0;') + prefix = [('_aux', '->', self)] + aux_ptr = 'xcb_out' + + elif context in ('unserialize', 'unpack'): + _c(' char *xcb_tmp = (char *)_buffer;') + if not self.is_switch: + if not self.var_followed_by_fixed_fields: + _c(' const %s *_aux = (%s *)_buffer;', self.c_type, self.c_type) + prefix = [('_aux', '->', self)] + else: + _c(' %s xcb_out;', self.c_type) + prefix = [('xcb_out', '.', self)] + else: + aux_var = '_aux' # default for unpack: single pointer + # note: unserialize not generated for switch + if 'unserialize' == context: + aux_var = '(*_aux)' # unserialize: double pointer (!) + prefix = [(aux_var, '->', self)] + aux_ptr = '*_aux' + _c(' unsigned int xcb_buffer_len = 0;') + _c(' unsigned int xcb_block_len = 0;') + _c(' unsigned int xcb_pad = 0;') + + elif 'sizeof' == context: + param_names = [p[2] for p in params] + if self.is_switch: + # switch: call _unpack() + _c(' %s _aux;', self.c_type) + _c(' return %s(%s, &_aux);', self.c_unpack_name, reduce(lambda x,y: "%s, %s" % (x, y), param_names)) + _c('}') + return + elif self.var_followed_by_fixed_fields: + # special case: call _unserialize() + _c(' return %s(%s, NULL);', self.c_unserialize_name, reduce(lambda x,y: "%s, %s" % (x, y), param_names)) + _c('}') + return + else: + _c(' char *xcb_tmp = (char *)_buffer;') + prefix = [('_aux', '->', self)] + + count = _c_serialize_helper(context, self, code_lines, temp_vars, prefix=prefix) + # update variable size fields (only important for context=='serialize' + variable_size_fields = count + if 'serialize' == context: + temp_vars.append(' unsigned int xcb_pad = 0;') + temp_vars.append(' char xcb_pad0[3] = {0, 0, 0};') + temp_vars.append(' struct iovec xcb_parts[%d];' % count) + temp_vars.append(' unsigned int xcb_parts_idx = 0;') + temp_vars.append(' unsigned int xcb_block_len = 0;') + temp_vars.append(' unsigned int i;') + temp_vars.append(' char *xcb_tmp;') + elif 'sizeof' == context: + # neither switch nor intermixed fixed and variable size fields: + # evaluate parameters directly + if not (self.is_switch or self.var_followed_by_fixed_fields): + + # look if we have to declare an '_aux' variable at all + if len(filter(lambda x: x.find('_aux')!=-1, code_lines))>0: + if not self.var_followed_by_fixed_fields: + _c(' const %s *_aux = (%s *)_buffer;', self.c_type, self.c_type) + else: + _c(' %s *_aux = malloc(sizeof(%s));', self.c_type, self.c_type) + + _c(' unsigned int xcb_buffer_len = 0;') + _c(' unsigned int xcb_block_len = 0;') + _c(' unsigned int xcb_pad = 0;') + + _c('') + for t in temp_vars: + _c(t) + _c('') + for l in code_lines: + _c(l) + + # variable sized fields have been collected, now + # allocate memory and copy everything into a continuous memory area + # note: this is not necessary in case of unpack + if context in ('serialize', 'unserialize'): + # unserialize: check for sizeof-only invocation + if 'unserialize' == context: + _c('') + _c(' if (NULL == _aux)') + _c(' return xcb_buffer_len;') + + _c('') + _c(' if (NULL == %s) {', aux_ptr) + _c(' /* allocate memory */') + _c(' %s = malloc(xcb_buffer_len);', aux_ptr) + if 'serialize' == context: + _c(' *_buffer = xcb_out;') + _c(' }') + _c('') + + # serialize: handle variable size fields in a loop + if 'serialize' == context: + if not self.is_switch and not self.var_followed_by_fixed_fields: + if len(wire_fields)>0: + _c(' *xcb_out = *_aux;') + # copy variable size fields into the buffer + if variable_size_fields > 0: + # xcb_out padding + if not self.is_switch and not self.var_followed_by_fixed_fields: + _c(' xcb_tmp = (char*)++xcb_out;') + _c(' xcb_tmp += xcb_out_pad;') + else: + _c(' xcb_tmp = xcb_out;') + + # variable sized fields + _c(' for(i=0; i<xcb_parts_idx; i++) {') + _c(' if (0 != xcb_parts[i].iov_base && 0 != xcb_parts[i].iov_len)') + _c(' memcpy(xcb_tmp, xcb_parts[i].iov_base, xcb_parts[i].iov_len);') + _c(' if (0 != xcb_parts[i].iov_len)') + _c(' xcb_tmp += xcb_parts[i].iov_len;') + _c(' }') + + # unserialize: assign variable size fields individually + if 'unserialize' == context: + _c(' xcb_tmp = ((char *)*_aux)+xcb_buffer_len;') + param_fields.reverse() + for field in param_fields: + if not field.type.fixed_size(): + _c(' xcb_tmp -= %s_len;', field.c_field_name) + _c(' memmove(xcb_tmp, %s, %s_len);', field.c_field_name, field.c_field_name) + _c(' *%s = xcb_out;', aux_ptr) + + _c('') + _c(' return xcb_buffer_len;') + _c('}') +# _c_serialize() + +def _c_iterator_get_end(field, accum): + ''' + Figures out what C code is needed to find the end of a variable-length structure field. + For nested structures, recurses into its last variable-sized field. + For lists, calls the end function + ''' + if field.type.is_container: + accum = field.c_accessor_name + '(' + accum + ')' + return _c_iterator_get_end(field.type.last_varsized_field, accum) + if field.type.is_list: + # XXX we can always use the first way + if field.type.member.is_simple: + return field.c_end_name + '(' + accum + ')' + else: + return field.type.member.c_end_name + '(' + field.c_iterator_name + '(' + accum + '))' + +def _c_iterator(self, name): + ''' + Declares the iterator structure and next/end functions for a given type. + ''' + _h_setlevel(0) + _h('') + _h('/**') + _h(' * @brief %s', self.c_iterator_type) + _h(' **/') + _h('typedef struct %s {', self.c_iterator_type) + _h(' %s *data; /**< */', self.c_type) + _h(' int%s rem; /**< */', ' ' * (len(self.c_type) - 2)) + _h(' int%s index; /**< */', ' ' * (len(self.c_type) - 2)) + _h('} %s;', self.c_iterator_type) + + _h_setlevel(1) + _c_setlevel(1) + _h('') + _h('/**') + _h(' * Get the next element of the iterator') + _h(' * @param i Pointer to a %s', self.c_iterator_type) + _h(' *') + _h(' * Get the next element in the iterator. The member rem is') + _h(' * decreased by one. The member data points to the next') + _h(' * element. The member index is increased by sizeof(%s)', self.c_type) + _h(' */') + _c('') + _hc('') + _hc('/*****************************************************************************') + _hc(' **') + _hc(' ** void %s', self.c_next_name) + _hc(' ** ') + _hc(' ** @param %s *i', self.c_iterator_type) + _hc(' ** @returns void') + _hc(' **') + _hc(' *****************************************************************************/') + _hc(' ') + _hc('void') + _h('%s (%s *i /**< */);', self.c_next_name, self.c_iterator_type) + _c('%s (%s *i /**< */)', self.c_next_name, self.c_iterator_type) + _c('{') + + if not self.fixed_size(): + _c(' %s *R = i->data;', self.c_type) + + if self.is_union: + # FIXME - how to determine the size of a variable size union?? + _c(' /* FIXME - determine the size of the union %s */', self.c_type) + else: + if self.need_sizeof: + _c(' xcb_generic_iterator_t child;') + _c(' child.data = (%s *)(((char *)R) + %s(R));', + self.c_type, self.c_sizeof_name) + _c(' i->index = (char *) child.data - (char *) i->data;') + else: + _c(' xcb_generic_iterator_t child = %s;', _c_iterator_get_end(self.last_varsized_field, 'R')) + _c(' i->index = child.index;') + _c(' --i->rem;') + _c(' i->data = (%s *) child.data;', self.c_type) + + else: + _c(' --i->rem;') + _c(' ++i->data;') + _c(' i->index += sizeof(%s);', self.c_type) + + _c('}') + + _h('') + _h('/**') + _h(' * Return the iterator pointing to the last element') + _h(' * @param i An %s', self.c_iterator_type) + _h(' * @return The iterator pointing to the last element') + _h(' *') + _h(' * Set the current element in the iterator to the last element.') + _h(' * The member rem is set to 0. The member data points to the') + _h(' * last element.') + _h(' */') + _c('') + _hc('') + _hc('/*****************************************************************************') + _hc(' **') + _hc(' ** xcb_generic_iterator_t %s', self.c_end_name) + _hc(' ** ') + _hc(' ** @param %s i', self.c_iterator_type) + _hc(' ** @returns xcb_generic_iterator_t') + _hc(' **') + _hc(' *****************************************************************************/') + _hc(' ') + _hc('xcb_generic_iterator_t') + _h('%s (%s i /**< */);', self.c_end_name, self.c_iterator_type) + _c('%s (%s i /**< */)', self.c_end_name, self.c_iterator_type) + _c('{') + _c(' xcb_generic_iterator_t ret;') + + if self.fixed_size(): + _c(' ret.data = i.data + i.rem;') + _c(' ret.index = i.index + ((char *) ret.data - (char *) i.data);') + _c(' ret.rem = 0;') + else: + _c(' while(i.rem > 0)') + _c(' %s(&i);', self.c_next_name) + _c(' ret.data = i.data;') + _c(' ret.rem = i.rem;') + _c(' ret.index = i.index;') + + _c(' return ret;') + _c('}') + +def _c_accessor_get_length(expr, field_mapping=None): + ''' + Figures out what C code is needed to get a length field. + The field_mapping parameter can be used to change the absolute name of a length field. + For fields that follow a variable-length field, use the accessor. + Otherwise, just reference the structure field directly. + ''' + + lenfield_name = expr.lenfield_name + if lenfield_name is not None: + if field_mapping is not None: + lenfield_name = field_mapping[lenfield_name][0] + + if expr.lenfield is not None and expr.lenfield.prev_varsized_field is not None: + # special case: variable and fixed size fields are intermixed + # if the lenfield is among the fixed size fields, there is no need + # to call a special accessor function like <expr.lenfield.c_accessor_name + '(' + prefix + ')'> + return field_mapping(expr.lenfield_name) + elif expr.lenfield_name is not None: + return lenfield_name + else: + return str(expr.nmemb) + +def _c_accessor_get_expr(expr, field_mapping): + ''' + Figures out what C code is needed to get the length of a list field. + The field_mapping parameter can be used to change the absolute name of a length field. + Recurses for math operations. + Returns bitcount for value-mask fields. + Otherwise, uses the value of the length field. + ''' + lenexp = _c_accessor_get_length(expr, field_mapping) + + if expr.op == '~': + return '(' + '~' + _c_accessor_get_expr(expr.rhs, field_mapping) + ')' + elif expr.op == 'popcount': + return 'xcb_popcount(' + _c_accessor_get_expr(expr.rhs, field_mapping) + ')' + elif expr.op == 'enumref': + enum_name = expr.lenfield_type.name + constant_name = expr.lenfield_name + c_name = _n(enum_name + (constant_name,)).upper() + return c_name + elif expr.op == 'sumof': + # locate the referenced list object + list_obj = expr.lenfield_type + field = None + for f in expr.lenfield_parent.fields: + if f.field_name == expr.lenfield_name: + field = f + break + + if field is None: + raise Exception("list field '%s' referenced by sumof not found" % expr.lenfield_name) + list_name = field_mapping[field.c_field_name][0] + c_length_func = "%s(%s)" % (field.c_length_name, list_name) + # note: xcb_sumof() has only been defined for integers + c_length_func = _c_accessor_get_expr(field.type.expr, field_mapping) + return 'xcb_sumof(%s, %s)' % (list_name, c_length_func) + elif expr.op != None: + return ('(' + _c_accessor_get_expr(expr.lhs, field_mapping) + + ' ' + expr.op + ' ' + + _c_accessor_get_expr(expr.rhs, field_mapping) + ')') + elif expr.bitfield: + return 'xcb_popcount(' + lenexp + ')' + else: + return lenexp + +def type_pad_type(type): + if type == 'void': + return 'char' + return type + +def _c_accessors_field(self, field): + ''' + Declares the accessor functions for a non-list field that follows a variable-length field. + ''' + c_type = self.c_type + + # special case: switch + switch_obj = self if self.is_switch else None + if self.is_bitcase: + switch_obj = self.parents[-1] + if switch_obj is not None: + c_type = switch_obj.c_type + + if field.type.is_simple: + _hc('') + _hc('') + _hc('/*****************************************************************************') + _hc(' ** ') + _hc(' ** %s %s', field.c_field_type, field.c_accessor_name) + _hc(' ** ') + _hc(' ** @param const %s *R', c_type) + _hc(' ** @returns %s', field.c_field_type) + _hc(' **') + _hc(' *****************************************************************************/') + _hc(' ') + _hc('%s', field.c_field_type) + _h('%s (const %s *R /**< */);', field.c_accessor_name, c_type) + _c('%s (const %s *R /**< */)', field.c_accessor_name, c_type) + _c('{') + if field.prev_varsized_field is None: + _c(' return (%s *) (R + 1);', field.c_field_type) + else: + _c(' xcb_generic_iterator_t prev = %s;', _c_iterator_get_end(field.prev_varsized_field, 'R')) + _c(' return * (%s *) ((char *) prev.data + XCB_TYPE_PAD(%s, prev.index) + %d);', + field.c_field_type, type_pad_type(field.first_field_after_varsized.type.c_type), field.prev_varsized_offset) + _c('}') + else: + _hc('') + _hc('') + _hc('/*****************************************************************************') + _hc(' **') + _hc(' ** %s * %s', field.c_field_type, field.c_accessor_name) + _hc(' ** ') + _hc(' ** @param const %s *R', c_type) + _hc(' ** @returns %s *', field.c_field_type) + _hc(' **') + _hc(' *****************************************************************************/') + _hc(' ') + if field.type.is_switch and switch_obj is None: + return_type = 'void *' + else: + return_type = '%s *' % field.c_field_type + + _hc(return_type) + _h('%s (const %s *R /**< */);', field.c_accessor_name, c_type) + _c('%s (const %s *R /**< */)', field.c_accessor_name, c_type) + _c('{') + if field.prev_varsized_field is None: + _c(' return (%s) (R + 1);', return_type) + # note: the special case 'variable fields followed by fixed size fields' + # is not of any consequence here, since the ordering gets + # 'corrected' in the reply function + else: + _c(' xcb_generic_iterator_t prev = %s;', _c_iterator_get_end(field.prev_varsized_field, 'R')) + _c(' return (%s) ((char *) prev.data + XCB_TYPE_PAD(%s, prev.index) + %d);', + return_type, type_pad_type(field.first_field_after_varsized.type.c_type), field.prev_varsized_offset) + _c('}') + + +def _c_accessors_list(self, field): + ''' + Declares the accessor functions for a list field. + Declares a direct-accessor function only if the list members are fixed size. + Declares length and get-iterator functions always. + ''' + list = field.type + c_type = self.c_type + + # special case: switch + # in case of switch, 2 params have to be supplied to certain accessor functions: + # 1. the anchestor object (request or reply) + # 2. the (anchestor) switch object + # the reason is that switch is either a child of a request/reply or nested in another switch, + # so whenever we need to access a length field, we might need to refer to some anchestor type + switch_obj = self if self.is_switch else None + if self.is_bitcase: + switch_obj = self.parents[-1] + if switch_obj is not None: + c_type = switch_obj.c_type + + params = [] + fields = {} + parents = self.parents if hasattr(self, 'parents') else [self] + # 'R': parents[0] is always the 'toplevel' container type + params.append(('const %s *R' % parents[0].c_type, parents[0])) + fields.update(_c_helper_field_mapping(parents[0], [('R', '->', parents[0])], flat=True)) + # auxiliary object for 'R' parameters + R_obj = parents[0] + + if switch_obj is not None: + # now look where the fields are defined that are needed to evaluate + # the switch expr, and store the parent objects in accessor_params and + # the fields in switch_fields + + # 'S': name for the 'toplevel' switch + toplevel_switch = parents[1] + params.append(('const %s *S' % toplevel_switch.c_type, toplevel_switch)) + fields.update(_c_helper_field_mapping(toplevel_switch, [('S', '->', toplevel_switch)], flat=True)) + + # initialize prefix for everything "below" S + prefix_str = '/* %s */ S' % toplevel_switch.name[-1] + prefix = [(prefix_str, '->', toplevel_switch)] + + # look for fields in the remaining containers + for p in parents[2:] + [self]: + # the separator between parent and child is always '.' here, + # because of nested switch statements + if not p.is_bitcase or (p.is_bitcase and p.has_name): + prefix.append((p.name[-1], '.', p)) + fields.update(_c_helper_field_mapping(p, prefix, flat=True)) + + # auxiliary object for 'S' parameter + S_obj = parents[1] + + _h_setlevel(1) + _c_setlevel(1) + if list.member.fixed_size(): + idx = 1 if switch_obj is not None else 0 + _hc('') + _hc('') + _hc('/*****************************************************************************') + _hc(' **') + _hc(' ** %s * %s', field.c_field_type, field.c_accessor_name) + _hc(' ** ') + _hc(' ** @param %s', params[idx][0]) + _hc(' ** @returns %s *', field.c_field_type) + _hc(' **') + _hc(' *****************************************************************************/') + _hc(' ') + _hc('%s *', field.c_field_type) + + _h('%s (%s /**< */);', field.c_accessor_name, params[idx][0]) + _c('%s (%s /**< */)', field.c_accessor_name, params[idx][0]) + + _c('{') + if switch_obj is not None: + _c(' return %s;', fields[field.c_field_name][0]) + elif field.prev_varsized_field is None: + _c(' return (%s *) (R + 1);', field.c_field_type) + else: + _c(' xcb_generic_iterator_t prev = %s;', _c_iterator_get_end(field.prev_varsized_field, 'R')) + _c(' return (%s *) ((char *) prev.data + XCB_TYPE_PAD(%s, prev.index) + %d);', + field.c_field_type, type_pad_type(field.first_field_after_varsized.type.c_type), field.prev_varsized_offset) + _c('}') + + _hc('') + _hc('') + _hc('/*****************************************************************************') + _hc(' **') + _hc(' ** int %s', field.c_length_name) + _hc(' ** ') + _hc(' ** @param const %s *R', c_type) + _hc(' ** @returns int') + _hc(' **') + _hc(' *****************************************************************************/') + _hc(' ') + _hc('int') + if switch_obj is not None: + _hc('%s (const %s *R /**< */,', field.c_length_name, R_obj.c_type) + spacing = ' '*(len(field.c_length_name)+2) + _h('%sconst %s *S /**< */);', spacing, S_obj.c_type) + _c('%sconst %s *S /**< */)', spacing, S_obj.c_type) + length = _c_accessor_get_expr(field.type.expr, fields) + else: + _h('%s (const %s *R /**< */);', field.c_length_name, c_type) + _c('%s (const %s *R /**< */)', field.c_length_name, c_type) + length = _c_accessor_get_expr(field.type.expr, fields) + _c('{') + _c(' return %s;', length) + _c('}') + + if field.type.member.is_simple: + _hc('') + _hc('') + _hc('/*****************************************************************************') + _hc(' **') + _hc(' ** xcb_generic_iterator_t %s', field.c_end_name) + _hc(' ** ') + _hc(' ** @param const %s *R', c_type) + _hc(' ** @returns xcb_generic_iterator_t') + _hc(' **') + _hc(' *****************************************************************************/') + _hc(' ') + _hc('xcb_generic_iterator_t') + if switch_obj is not None: + _hc('%s (const %s *R /**< */,', field.c_end_name, R_obj.c_type) + spacing = ' '*(len(field.c_end_name)+2) + _h('%sconst %s *S /**< */);', spacing, S_obj.c_type) + _c('%sconst %s *S /**< */)', spacing, S_obj.c_type) + else: + _h('%s (const %s *R /**< */);', field.c_end_name, c_type) + _c('%s (const %s *R /**< */)', field.c_end_name, c_type) + _c('{') + _c(' xcb_generic_iterator_t i;') + + param = 'R' if switch_obj is None else 'S' + if switch_obj is not None: + _c(' i.data = %s + %s;', fields[field.c_field_name][0], + _c_accessor_get_expr(field.type.expr, fields)) + elif field.prev_varsized_field == None: + _c(' i.data = ((%s *) (R + 1)) + (%s);', field.type.c_wiretype, + _c_accessor_get_expr(field.type.expr, fields)) + else: + _c(' xcb_generic_iterator_t child = %s;', + _c_iterator_get_end(field.prev_varsized_field, 'R')) + _c(' i.data = ((%s *) child.data) + (%s);', field.type.c_wiretype, + _c_accessor_get_expr(field.type.expr, fields)) + + _c(' i.rem = 0;') + _c(' i.index = (char *) i.data - (char *) %s;', param) + _c(' return i;') + _c('}') + + else: + _hc('') + _hc('') + _hc('/*****************************************************************************') + _hc(' **') + _hc(' ** %s %s', field.c_iterator_type, field.c_iterator_name) + _hc(' ** ') + _hc(' ** @param const %s *R', c_type) + _hc(' ** @returns %s', field.c_iterator_type) + _hc(' **') + _hc(' *****************************************************************************/') + _hc(' ') + + _hc('%s', field.c_iterator_type) + if switch_obj is not None: + _hc('%s (const %s *R /**< */,', field.c_iterator_name, R_obj.c_type) + spacing = ' '*(len(field.c_iterator_name)+2) + _h('%sconst %s *S /**< */);', spacing, S_obj.c_type) + _c('%sconst %s *S /**< */)', spacing, S_obj.c_type) + else: + _h('%s (const %s *R /**< */);', field.c_iterator_name, c_type) + _c('%s (const %s *R /**< */)', field.c_iterator_name, c_type) + _c('{') + _c(' %s i;', field.c_iterator_type) + + if switch_obj is not None: + _c(' i.data = %s;', fields[field.c_field_name][0]) + _c(' i.rem = %s;', _c_accessor_get_expr(field.type.expr, fields)) + elif field.prev_varsized_field == None: + _c(' i.data = (%s *) (R + 1);', field.c_field_type) + else: + _c(' xcb_generic_iterator_t prev = %s;', _c_iterator_get_end(field.prev_varsized_field, 'R')) + _c(' i.data = (%s *) ((char *) prev.data + XCB_TYPE_PAD(%s, prev.index));', + field.c_field_type, type_pad_type(field.c_field_type)) + if switch_obj is None: + _c(' i.rem = %s;', _c_accessor_get_expr(field.type.expr, fields)) + _c(' i.index = (char *) i.data - (char *) %s;', 'R' if switch_obj is None else 'S' ) + _c(' return i;') + _c('}') + +def _c_accessors(self, name, base): + ''' + Declares the accessor functions for the fields of a structure. + ''' + # no accessors for switch itself - + # switch always needs to be unpacked explicitly +# if self.is_switch: +# pass +# else: + if True: + for field in self.fields: + if field.type.is_list and not field.type.fixed_size(): + _c_accessors_list(self, field) + elif field.prev_varsized_field is not None or not field.type.fixed_size(): + _c_accessors_field(self, field) + +def c_simple(self, name): + ''' + Exported function that handles cardinal type declarations. + These are types which are typedef'd to one of the CARDx's, char, float, etc. + ''' + _c_type_setup(self, name, ()) + + if (self.name != name): + # Typedef + _h_setlevel(0) + my_name = _t(name) + _h('') + _h('typedef %s %s;', _t(self.name), my_name) + + # Iterator + _c_iterator(self, name) + +def _c_complex(self): + ''' + Helper function for handling all structure types. + Called for all structs, requests, replies, events, errors. + ''' + _h_setlevel(0) + _h('') + _h('/**') + _h(' * @brief %s', self.c_type) + _h(' **/') + _h('typedef %s %s {', self.c_container, self.c_type) + + struct_fields = [] + maxtypelen = 0 + + varfield = None + for field in self.fields: + if not field.type.fixed_size() and not self.is_switch and not self.is_union: + varfield = field.c_field_name + continue + if field.wire: + struct_fields.append(field) + + for field in struct_fields: + length = len(field.c_field_type) + # account for '*' pointer_spec + if not field.type.fixed_size(): + length += 1 + maxtypelen = max(maxtypelen, length) + + def _c_complex_field(self, field, space=''): + if (field.type.fixed_size() or + # in case of switch with switch children, don't make the field a pointer + # necessary for unserialize to work + (self.is_switch and field.type.is_switch)): + spacing = ' ' * (maxtypelen - len(field.c_field_type)) + _h('%s %s%s %s%s; /**< */', space, field.c_field_type, spacing, field.c_field_name, field.c_subscript) + else: + spacing = ' ' * (maxtypelen - (len(field.c_field_type) + 1)) + _h('%s %s%s *%s%s; /**< */', space, field.c_field_type, spacing, field.c_field_name, field.c_subscript) + + if not self.is_switch: + for field in struct_fields: + _c_complex_field(self, field) + else: + for b in self.bitcases: + space = '' + if b.type.has_name: + _h(' struct _%s {', b.c_field_name) + space = ' ' + for field in b.type.fields: + _c_complex_field(self, field, space) + if b.type.has_name: + _h(' } %s;', b.c_field_name) + + _h('} %s;', self.c_type) + +def c_struct(self, name): + ''' + Exported function that handles structure declarations. + ''' + _c_type_setup(self, name, ()) + _c_complex(self) + _c_accessors(self, name, name) + _c_iterator(self, name) + +def c_union(self, name): + ''' + Exported function that handles union declarations. + ''' + _c_type_setup(self, name, ()) + _c_complex(self) + _c_iterator(self, name) + +def _c_request_helper(self, name, cookie_type, void, regular, aux=False): + ''' + Declares a request function. + ''' + + # Four stunningly confusing possibilities here: + # + # Void Non-void + # ------------------------------ + # "req" "req" + # 0 flag CHECKED flag Normal Mode + # void_cookie req_cookie + # ------------------------------ + # "req_checked" "req_unchecked" + # CHECKED flag 0 flag Abnormal Mode + # void_cookie req_cookie + # ------------------------------ + + + # Whether we are _checked or _unchecked + checked = void and not regular + unchecked = not void and not regular + + # What kind of cookie we return + func_cookie = 'xcb_void_cookie_t' if void else self.c_cookie_type + + # What flag is passed to xcb_request + func_flags = '0' if (void and regular) or (not void and not regular) else 'XCB_REQUEST_CHECKED' + + # Global extension id variable or NULL for xproto + func_ext_global = '&' + _ns.c_ext_global_name if _ns.is_ext else '0' + + # What our function name is + func_name = self.c_request_name if not aux else self.c_aux_name + if checked: + func_name = self.c_checked_name if not aux else self.c_aux_checked_name + if unchecked: + func_name = self.c_unchecked_name if not aux else self.c_aux_unchecked_name + + param_fields = [] + wire_fields = [] + maxtypelen = len('xcb_connection_t') + serial_fields = [] + # special case: list with variable size elements + list_with_var_size_elems = False + + for field in self.fields: + if field.visible: + # The field should appear as a call parameter + param_fields.append(field) + if field.wire and not field.auto: + # We need to set the field up in the structure + wire_fields.append(field) + if field.type.need_serialize or field.type.need_sizeof: + serial_fields.append(field) + + for field in param_fields: + c_field_const_type = field.c_field_const_type + if field.type.need_serialize and not aux: + c_field_const_type = "const void" + if len(c_field_const_type) > maxtypelen: + maxtypelen = len(c_field_const_type) + if field.type.is_list and not field.type.member.fixed_size(): + list_with_var_size_elems = True + + _h_setlevel(1) + _c_setlevel(1) + _h('') + _h('/**') + _h(' * Delivers a request to the X server') + _h(' * @param c The connection') + _h(' * @return A cookie') + _h(' *') + _h(' * Delivers a request to the X server.') + _h(' * ') + if checked: + _h(' * This form can be used only if the request will not cause') + _h(' * a reply to be generated. Any returned error will be') + _h(' * saved for handling by xcb_request_check().') + if unchecked: + _h(' * This form can be used only if the request will cause') + _h(' * a reply to be generated. Any returned error will be') + _h(' * placed in the event queue.') + _h(' */') + _c('') + _hc('') + _hc('/*****************************************************************************') + _hc(' **') + _hc(' ** %s %s', cookie_type, func_name) + _hc(' ** ') + + spacing = ' ' * (maxtypelen - len('xcb_connection_t')) + _hc(' ** @param xcb_connection_t%s *c', spacing) + + for field in param_fields: + c_field_const_type = field.c_field_const_type + if field.type.need_serialize and not aux: + c_field_const_type = "const void" + spacing = ' ' * (maxtypelen - len(c_field_const_type)) + _hc(' ** @param %s%s %s%s', c_field_const_type, spacing, field.c_pointer, field.c_field_name) + + _hc(' ** @returns %s', cookie_type) + _hc(' **') + _hc(' *****************************************************************************/') + _hc(' ') + _hc('%s', cookie_type) + + spacing = ' ' * (maxtypelen - len('xcb_connection_t')) + comma = ',' if len(param_fields) else ');' + _h('%s (xcb_connection_t%s *c /**< */%s', func_name, spacing, comma) + comma = ',' if len(param_fields) else ')' + _c('%s (xcb_connection_t%s *c /**< */%s', func_name, spacing, comma) + + func_spacing = ' ' * (len(func_name) + 2) + count = len(param_fields) + for field in param_fields: + count = count - 1 + c_field_const_type = field.c_field_const_type + c_pointer = field.c_pointer + if field.type.need_serialize and not aux: + c_field_const_type = "const void" + c_pointer = '*' + spacing = ' ' * (maxtypelen - len(c_field_const_type)) + comma = ',' if count else ');' + _h('%s%s%s %s%s /**< */%s', func_spacing, c_field_const_type, + spacing, c_pointer, field.c_field_name, comma) + comma = ',' if count else ')' + _c('%s%s%s %s%s /**< */%s', func_spacing, c_field_const_type, + spacing, c_pointer, field.c_field_name, comma) + + count = 2 + if not self.var_followed_by_fixed_fields: + for field in param_fields: + if not field.type.fixed_size(): + count = count + 2 + if field.type.need_serialize: + # _serialize() keeps track of padding automatically + count -= 1 + dimension = count + 2 + + _c('{') + _c(' static const xcb_protocol_request_t xcb_req = {') + _c(' /* count */ %d,', count) + _c(' /* ext */ %s,', func_ext_global) + _c(' /* opcode */ %s,', self.c_request_name.upper()) + _c(' /* isvoid */ %d', 1 if void else 0) + _c(' };') + _c(' ') + + _c(' struct iovec xcb_parts[%d];', dimension) + _c(' %s xcb_ret;', func_cookie) + _c(' %s xcb_out;', self.c_type) + if self.var_followed_by_fixed_fields: + _c(' /* in the protocol description, variable size fields are followed by fixed size fields */') + _c(' void *xcb_aux = 0;') + + + for idx, f in enumerate(serial_fields): + if aux: + _c(' void *xcb_aux%d = 0;' % (idx)) + if list_with_var_size_elems: + _c(' unsigned int i;') + _c(' unsigned int xcb_tmp_len;') + _c(' char *xcb_tmp;') + _c(' ') + # simple request call tracing +# _c(' printf("in function %s\\n");' % func_name) + + # fixed size fields + for field in wire_fields: + if field.type.fixed_size(): + if field.type.is_expr: + _c(' xcb_out.%s = %s;', field.c_field_name, _c_accessor_get_expr(field.type.expr, None)) + elif field.type.is_pad: + if field.type.nmemb == 1: + _c(' xcb_out.%s = 0;', field.c_field_name) + else: + _c(' memset(xcb_out.%s, 0, %d);', field.c_field_name, field.type.nmemb) + else: + if field.type.nmemb == 1: + _c(' xcb_out.%s = %s;', field.c_field_name, field.c_field_name) + else: + _c(' memcpy(xcb_out.%s, %s, %d);', field.c_field_name, field.c_field_name, field.type.nmemb) + + def get_serialize_args(type_obj, c_field_name, aux_var, context='serialize'): + serialize_args = get_serialize_params(context, type_obj, + c_field_name, + aux_var)[2] + return reduce(lambda x,y: "%s, %s" % (x,y), [a[2] for a in serialize_args]) + + # calls in order to free dyn. all. memory + free_calls = [] + + _c(' ') + if not self.var_followed_by_fixed_fields: + _c(' xcb_parts[2].iov_base = (char *) &xcb_out;') + _c(' xcb_parts[2].iov_len = sizeof(xcb_out);') + _c(' xcb_parts[3].iov_base = 0;') + _c(' xcb_parts[3].iov_len = -xcb_parts[2].iov_len & 3;') + + count = 4 + + for field in param_fields: + if not field.type.fixed_size(): + _c(' /* %s %s */', field.type.c_type, field.c_field_name) + # default: simple cast to char * + if not field.type.need_serialize and not field.type.need_sizeof: + _c(' xcb_parts[%d].iov_base = (char *) %s;', count, field.c_field_name) + if field.type.is_list: + if field.type.member.fixed_size(): + _c(' xcb_parts[%d].iov_len = %s * sizeof(%s);', count, + _c_accessor_get_expr(field.type.expr, None), + field.type.member.c_wiretype) + else: + list_length = _c_accessor_get_expr(field.type.expr, None) + + length = '' + _c(" xcb_parts[%d].iov_len = 0;" % count) + _c(" xcb_tmp = (char *)%s;", field.c_field_name) + _c(" for(i=0; i<%s; i++) {" % list_length) + _c(" xcb_tmp_len = %s(xcb_tmp);" % + (field.type.c_sizeof_name)) + _c(" xcb_parts[%d].iov_len += xcb_tmp_len;" % count) + _c(" xcb_tmp += xcb_tmp_len;") + _c(" }") + else: + # not supposed to happen + raise Exception("unhandled variable size field %s" % field.c_field_name) + else: + if not aux: + _c(' xcb_parts[%d].iov_base = (char *) %s;', count, field.c_field_name) + idx = serial_fields.index(field) + aux_var = '&xcb_aux%d' % idx + context = 'serialize' if aux else 'sizeof' + _c(' xcb_parts[%d].iov_len = ', count) + if aux: + serialize_args = get_serialize_args(field.type, aux_var, field.c_field_name, context) + _c(' %s (%s);', field.type.c_serialize_name, serialize_args) + _c(' xcb_parts[%d].iov_base = xcb_aux%d;' % (count, idx)) + free_calls.append(' free(xcb_aux%d);' % idx) + else: + serialize_args = get_serialize_args(field.type, field.c_field_name, aux_var, context) + func_name = field.type.c_sizeof_name + _c(' %s (%s);', func_name, serialize_args) + + count += 1 + if not (field.type.need_serialize or field.type.need_sizeof): + # the _serialize() function keeps track of padding automatically + _c(' xcb_parts[%d].iov_base = 0;', count) + _c(' xcb_parts[%d].iov_len = -xcb_parts[%d].iov_len & 3;', count, count-1) + count += 1 + + # elif self.var_followed_by_fixed_fields: + else: + _c(' xcb_parts[2].iov_base = (char *) &xcb_out;') + # request header: opcodes + length + _c(' xcb_parts[2].iov_len = 2*sizeof(uint8_t) + sizeof(uint16_t);') + count += 1 + # call _serialize() + buffer_var = '&xcb_aux' + serialize_args = get_serialize_args(self, buffer_var, '&xcb_out', 'serialize') + _c(' xcb_parts[%d].iov_len = %s (%s);', count, self.c_serialize_name, serialize_args) + _c(' xcb_parts[%d].iov_base = (char *) xcb_aux;', count) + free_calls.append(' free(xcb_aux);') + # no padding necessary - _serialize() keeps track of padding automatically + + _c(' ') + _c(' xcb_ret.sequence = xcb_send_request(c, %s, xcb_parts + 2, &xcb_req);', func_flags) + + # free dyn. all. data, if any + for f in free_calls: + _c(f) + _c(' return xcb_ret;') + _c('}') + +def _c_reply(self, name): + ''' + Declares the function that returns the reply structure. + ''' + spacing1 = ' ' * (len(self.c_cookie_type) - len('xcb_connection_t')) + spacing2 = ' ' * (len(self.c_cookie_type) - len('xcb_generic_error_t')) + spacing3 = ' ' * (len(self.c_reply_name) + 2) + + # check if _unserialize() has to be called for any field + def look_for_special_cases(complex_obj): + unserialize_fields = [] + # no unserialize call in case of switch + if not complex_obj.is_switch: + for field in complex_obj.fields: + # three cases: 1. field with special case + # 2. container that contains special case field + # 3. list with special case elements + if field.type.var_followed_by_fixed_fields: + unserialize_fields.append(field) + elif field.type.is_container: + unserialize_fields += look_for_special_cases(field.type) + elif field.type.is_list: + if field.type.member.var_followed_by_fixed_fields: + unserialize_fields.append(field) + if field.type.member.is_container: + unserialize_fields += look_for_special_cases(field.type.member) + return unserialize_fields + + unserialize_fields = look_for_special_cases(self.reply) + + _h('') + _h('/**') + _h(' * Return the reply') + _h(' * @param c The connection') + _h(' * @param cookie The cookie') + _h(' * @param e The xcb_generic_error_t supplied') + _h(' *') + _h(' * Returns the reply of the request asked by') + _h(' * ') + _h(' * The parameter @p e supplied to this function must be NULL if') + _h(' * %s(). is used.', self.c_unchecked_name) + _h(' * Otherwise, it stores the error if any.') + _h(' *') + _h(' * The returned value must be freed by the caller using free().') + _h(' */') + _c('') + _hc('') + _hc('/*****************************************************************************') + _hc(' **') + _hc(' ** %s * %s', self.c_reply_type, self.c_reply_name) + _hc(' ** ') + _hc(' ** @param xcb_connection_t%s *c', spacing1) + _hc(' ** @param %s cookie', self.c_cookie_type) + _hc(' ** @param xcb_generic_error_t%s **e', spacing2) + _hc(' ** @returns %s *', self.c_reply_type) + _hc(' **') + _hc(' *****************************************************************************/') + _hc(' ') + _hc('%s *', self.c_reply_type) + _hc('%s (xcb_connection_t%s *c /**< */,', self.c_reply_name, spacing1) + _hc('%s%s cookie /**< */,', spacing3, self.c_cookie_type) + _h('%sxcb_generic_error_t%s **e /**< */);', spacing3, spacing2) + _c('%sxcb_generic_error_t%s **e /**< */)', spacing3, spacing2) + _c('{') + + if len(unserialize_fields)>0: + # certain variable size fields need to be unserialized explicitly + _c(' %s *reply = (%s *) xcb_wait_for_reply(c, cookie.sequence, e);', + self.c_reply_type, self.c_reply_type) + _c(' int i;') + for field in unserialize_fields: + if field.type.is_list: + _c(' %s %s_iter = %s(reply);', field.c_iterator_type, field.c_field_name, field.c_iterator_name) + _c(' int %s_len = %s(reply);', field.c_field_name, field.c_length_name) + _c(' %s *%s_data;', field.c_field_type, field.c_field_name) + else: + raise Exception('not implemented: call _unserialize() in reply for non-list type %s', field.c_field_type) + # call _unserialize(), using the reply as source and target buffer + _c(' /* special cases: transform parts of the reply to match XCB data structures */') + for field in unserialize_fields: + if field.type.is_list: + _c(' for(i=0; i<%s_len; i++) {', field.c_field_name) + _c(' %s_data = %s_iter.data;', field.c_field_name, field.c_field_name) + _c(' %s((const void *)%s_data, &%s_data);', field.type.c_unserialize_name, + field.c_field_name, field.c_field_name) + _c(' %s(&%s_iter);', field.type.c_next_name, field.c_field_name) + _c(' }') + # return the transformed reply + _c(' return reply;') + + else: + _c(' return (%s *) xcb_wait_for_reply(c, cookie.sequence, e);', self.c_reply_type) + + _c('}') + +def _c_opcode(name, opcode): + ''' + Declares the opcode define for requests, events, and errors. + ''' + _h_setlevel(0) + _h('') + _h('/** Opcode for %s. */', _n(name)) + _h('#define %s %s', _n(name).upper(), opcode) + +def _c_cookie(self, name): + ''' + Declares the cookie type for a non-void request. + ''' + _h_setlevel(0) + _h('') + _h('/**') + _h(' * @brief %s', self.c_cookie_type) + _h(' **/') + _h('typedef struct %s {', self.c_cookie_type) + _h(' unsigned int sequence; /**< */') + _h('} %s;', self.c_cookie_type) + +def c_request(self, name): + ''' + Exported function that handles request declarations. + ''' + _c_type_setup(self, name, ('request',)) + + if self.reply: + # Cookie type declaration + _c_cookie(self, name) + + # Opcode define + _c_opcode(name, self.opcode) + + # Request structure declaration + _c_complex(self) + + if self.reply: + _c_type_setup(self.reply, name, ('reply',)) + # Reply structure definition + _c_complex(self.reply) + # Request prototypes + _c_request_helper(self, name, self.c_cookie_type, False, True) + _c_request_helper(self, name, self.c_cookie_type, False, False) + if self.need_aux: + _c_request_helper(self, name, self.c_cookie_type, False, True, True) + _c_request_helper(self, name, self.c_cookie_type, False, False, True) + # Reply accessors + _c_accessors(self.reply, name + ('reply',), name) + _c_reply(self, name) + else: + # Request prototypes + _c_request_helper(self, name, 'xcb_void_cookie_t', True, False) + _c_request_helper(self, name, 'xcb_void_cookie_t', True, True) + if self.need_aux: + _c_request_helper(self, name, 'xcb_void_cookie_t', True, False, True) + _c_request_helper(self, name, 'xcb_void_cookie_t', True, True, True) + + +def c_event(self, name): + ''' + Exported function that handles event declarations. + ''' + _c_type_setup(self, name, ('event',)) + + # Opcode define + _c_opcode(name, self.opcodes[name]) + + if self.name == name: + # Structure definition + _c_complex(self) + else: + # Typedef + _h('') + _h('typedef %s %s;', _t(self.name + ('event',)), _t(name + ('event',))) + +def c_error(self, name): + ''' + Exported function that handles error declarations. + ''' + _c_type_setup(self, name, ('error',)) + + # Opcode define + _c_opcode(name, self.opcodes[name]) + + if self.name == name: + # Structure definition + _c_complex(self) + else: + # Typedef + _h('') + _h('typedef %s %s;', _t(self.name + ('error',)), _t(name + ('error',))) + + +# Main routine starts here + +# Must create an "output" dictionary before any xcbgen imports. +output = {'open' : c_open, + 'close' : c_close, + 'simple' : c_simple, + 'enum' : c_enum, + 'struct' : c_struct, + 'union' : c_union, + 'request' : c_request, + 'event' : c_event, + 'error' : c_error, + } + +# Boilerplate below this point + +# Check for the argument that specifies path to the xcbgen python package. +try: + opts, args = getopt.getopt(sys.argv[1:], 'p:') +except getopt.GetoptError, err: + print str(err) + print 'Usage: c_client.py [-p path] file.xml' + sys.exit(1) + +for (opt, arg) in opts: + if opt == '-p': + sys.path.append(arg) + +# Import the module class +try: + from xcbgen.state import Module +except ImportError: + print '' + print 'Failed to load the xcbgen Python package!' + print 'Make sure that xcb/proto installed it on your Python path.' + print 'If not, you will need to create a .pth file or define $PYTHONPATH' + print 'to extend the path.' + print 'Refer to the README file in xcb/proto for more info.' + print '' + raise + +# Parse the xml header +module = Module(args[0], output) + +# Build type-registry and resolve type dependencies +module.register() +module.resolve() + +# Output the code +module.generate() |