Re: gdb/1222: fails to find correct source file for a generated Cfile with #line directives

[Britton <fsblk at aurora dot uaf dot edu>]

The following reply was made to PR symtab/1222; it has been noted by GNATS.

From: Britton <fsblk@aurora.uaf.edu>
To: Jim Blandy <jimb@redhat.com>
Cc: <gdb-gnats@sources.redhat.com>
Subject: Re: gdb/1222: fails to find correct source file for a generated C
 file with #line directives
Date: Wed, 21 May 2003 15:52:05 -0800 (AKDT)

 It was DWARF-2, and compiling with -gstabs does indeed fix the problem,
 thanks.  It still looks like a thing very worth fixing since autoconf
 and everything else tends to default to -g which apparently gives you
 DWARF-2.
 
 Britton Kerin
 __
 GNU GPL: "The Source will be with you... always."
 
 On 21 May 2003, Jim Blandy wrote:
 
 >
 > Which debug format are you using?  If you're using Dwarf 2, I think
 > this is a problem in GDB's Dwarf 2 reader.
 >
 > You can tell by running the command 'readelf -S foo.o' on your object
 > files.  You'll see entries like this:
 >
 >   [27] .debug_info       PROGBITS        00000000 27433f 8978bf 00  0   0  1
 >   [28] .debug_abbrev     PROGBITS        00000000 b0bbfe 0353e4 00  0   0  1
 >
 > If it contains Dwarf 2 info, and entries like this:
 >
 >   [23] .stab             PROGBITS        00000000 0004a4 000024 0c  24   0  4
 >   [24] .stabstr          STRTAB          00000000 0004c8 000036 00   0   0  1
 >
 > if it contains STABS.
 >
 > The problem is that the Dwarf 2 reader doesn't construct partial
 > symbol tables for anything but the main .c file; files referred to via
 > #line directives or #includes don't get recorded at all.  Of course,
 > if you do anything that causes GDB to read the full symbols for that
 > .c file, at that point you'll get symtabs for all the source files ---
 > including ones mentioned via #line.
 >
 > The solution would be for the Dwarf 2 reader to go look at the source
 > location info of each DW_TAG_compile_unit die, and build psymtabs for
 > all the filenames it finds there.
 >