- more strickling differentiation variable location between
a register, and the dereferenced address defined by a
register (and possibly an offset)
- added a parameter to symt_add_func_local (and internal symbol
data struct) to help differentiate
- fix all debug parsers to confer to this new scheme
- now internally storing module info as a 64bit module structure
(the interest in not in the 64 bit side of things, but because
it allows storing lots of usefull information)
- fixed SymGetModuleInfo64W which wasn't returning the
LoadedPdbName field
- now filling for MSC files with the extended information
(instead of guessing it)
- reused the extended module info to store wine ELF module
information (link for map link, stabs vs dwarf symbol info)
- added the elf_is_in_thunk_area() function to locate an address
within the known thunk area of Wine's builtin modules
- now passing this thunk information to the dwarf parser so that it
can drop functions from the thunk areas (as dwarf symbols), so that
those functions can be later on marked as thunks in dbghelp
internals
- fixed regression when looking for winebuild's special symbols in ELF symtab
- STT_NOTYPE is still accepted as a valid symbol type (on top of STT_FUNC and STT_OBJECT)
When several debug formats (eg stabs & dwarf2) exist in the same
module, allow to load them all (this is needed when different
compilation units are compiled with different debug options).
- split the data structures for walking through the data (abbrev,
cu... -> traverse_ctx) and data manipulation/loading (-> parse_ctx)
- created dwarf2_parse_compilation_unit() for clarity
- added section structure to gather information about all ELF sections
related to all dwarf2 sections (more to come)
- properly handling function blocks
- handling of variables inside of blocks (either on stack, or in a
register, or relative to the frame register if any)
- added dwarf2_subprogram_t structure so that we can pass around
relevant information for function parsing
- added the ability to tell to add_func_local whether we're adding
a local variable or a parameter (and removed the black magic
we were using)
- we can now address variables defined as an offset to a register
- various fixes to follow better the functions definition (nesting,
subroutines...)
- don't directly parse types referenced by a subprogram, they'll be
loaded anyway when needed
Andrew Talbot