This is the mail archive of the gdb-patches@sources.redhat.com mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[rfa] Deprecate msymbol.info, add msymbol.bfd_symbol?


There isn't yet a patch yet. I wan't to run the theory past people, especially the symtab maintainers.

A recent thread discussed the "info" field of the symbol table where MichaelC was trying to pack it down, while I was arguing that it should "just go", instead have the computation performed by MAKE_MSYMBOL_SPECIAL on every minimal-symbol performed on-demand (and after the symbo-table load).

The problem with eliminating "info" is that it would leave the on-demand code with no efficient way to find the underlying BFD info that it needes :-(

Consequently, I'd like to propose that "info" be superseeded by a "struct bfd_symbol *" pointer.

I say "superseed" as I'd prefer doing this in two steps. In addition to the per-architecture MAKE_MSYMBOL_SPECIAL, readers like elfread.c also squirrel away info in that field :-/

The consequence will be that, for a time, the minimal symbol table will be bigger. Since its the minimal-symbol table I don't feel too guilty, long term the minimal-symbol's size will be restored. In fact, given the apparent overlap between bfd_symbol and general_symbol_info (name, section, bfd_section, ?), it opens the way to further savings.

thoughts? does the theory look correct?
Andrew


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]