https://sourceware.org/bugzilla/show_bug.cgi?id=28266
--- Comment #3 from Joel Rosdahl <joel at rosdahl dot net> ---
(In reply to Nick Clifton from comment #2)
> So I would favour the interpretation that -fdebug-prefix-map is expected to
> override strict conformance to the DWARF standard because of the desire/need
> for reproducible builds.
Sounds good.
> Possibly it can be argued that in the presence of the -fdebug-prefix-map
> option the DW_AT_comp_dir entry ought to be changed as well. (Which would
> match the intention of creating reproducible builds).
>
> The problem with this idea is that it is GCC not GAS that generates the
> DW_AT_comp_dir entry, and detecting it reliably in the assembler would quite
> hard. But maybe GCC can be persuaded to do this instead ?
> [...]
> So gas and Clang are in sync on this ? Good.
Hmm, sounds like I probably was unclear in my original description.
DW_AT_comp_dir is already changed by -fdebug-prefix-map, but the 0'th directory
table entry is not (after 3417bfca676ff0b440f070a09837813e99b2c731).
Here is how to reproduce the issue:
% cd /tmp
% echo 'int x;' >test.c
% /usr/bin/gcc -g -fdebug-prefix-map=/tmp=/example -c test.c
% objdump -W test.o 2>/dev/null | grep -E '/tmp|/example'
<16> DW_AT_comp_dir : (indirect line string, offset: 0x0):
/example
0 (indirect line string, offset: 0x10): /tmp
1 (indirect line string, offset: 0x15): /example
0x00000000 2f657861 6d706c65 00746573 742e6300 /example.test.c.
0x00000010 2f746d70 002f6578 616d706c 65007465 /tmp./example.te
DW_AT_comp_dir is rewritten according to -fdebug-prefix-map but the 0'th
directory table entry is still "/tmp".
With Clang there is no trace of "/tmp":
% cd /tmp
% echo 'int x;' >test.c
% /usr/bin/clang -gdwarf-5 -fdebug-prefix-map=/tmp=/example -c test.c
% objdump -W test.o 2>/dev/null | grep -E '/tmp|/example'
2 33 /example
0x00000030 2e63002f 6578616d 706c6500 7800696e .c./example.x.in
0 (indirect line string, offset: 0x0): /example
0x00000000 2f657861 6d706c65 00746573 742e6300 /example.test.c.
--
You are receiving this mail because:
You are on the CC list for the bug.
