Hi!
This is a question regarding command-line options for GCC vs. binutils
assembler, linker. It came up during a patch review related to nvptx
assembler and linker, which live outside of binutils:
<https://github.com/MentorEmbedded/nvptx-tools>, and don't share the
standard binutils assembler, linker framework (because they're rather
different/simpler). As such they also don't share the standard binutils
'as', 'ld' command-line interface, though it's of course helpful (to
users) to be compatible, as much as is feasible.
I had proposed for GCC "nvptx: forward '-v' command-line option to
assembler, linker" (see snippets quoted below), and during review we
found:
On 2022-09-05T17:02:26+0200, Tom de Vries via Gcc-patches
<gcc-patc...@gcc.gnu.org> wrote:
> On 6/7/22 17:41, Thomas Schwinge wrote:
>> --- a/gcc/config/nvptx/nvptx.h
>> +++ b/gcc/config/nvptx/nvptx.h
>> +/* Assembler supports '-v' option; handle similar to
>> + '../../gcc.cc:asm_options', 'HAVE_GNU_AS'. */
>> +#define ASM_SPEC "%{v}"
> The ASM_SPEC part LGTM.
For reference, that GCC '-v' gets forwarded to binutils assembler is due
to 'gcc/gcc.cc:asm_options':
1272 #if HAVE_GNU_AS
1273 /* If GNU AS is used, then convert -w (no warnings), -I, and -v
1274 to the assembler equivalents. */
1275 "%{v} %{w:-W} %{I*} "
1276 #endif
That was added by Carlos some 15 years ago; Subversion r127275
(Git commit dc60b7754db561a029e42eeb297493726f8b8b33),
<https://inbox.sourceware.org/gcc-patches/20070801175134.GR18317@lios>
"Pass more options to the assembler". ('-w', '-I' are not applicable to
nvptx 'as'.)
But that was for binutils assembler only, not for binutils linker, so
when for nvptx I additionally proposed:
>> +/* Linker supports '-v' option. */
>> +#define LINK_SPEC "%{v}"
..., Tom rightfully asked:
> [...] I wonder, normally we don't pass -v to ld, and need -Wl,-v for
> that.
So, on my quest for making things uniform/simple, I now wonder: should we
also forward GCC '-v' to binutils linker, or is there a reason to not do
that?
(As far as I can tell, GCC 'collect2' is also already doing the right
thing regarding '-v'.)
I've manually run a few experiments with 'gcc -v -Wl,-v' (proposed to
then simplify to just 'gcc -v'), with both binutils 'ld.bfd', 'ld.gold',
and have not spotted any issues. As 'ld.mold', 'ld.lld' also identify as
"compatible with GNU ld", I've likewise tested them, and again not
spotted any issues. So I think we should be save to generally handle
'%{v}' linker spec for all 'HAVE_GNU_LD', or does anyone see any problem
with that?
(I've not yet thought about where in 'gcc/gcc.cc' to place '%{v}' linker
spec, guarded by '#if HAVE_GNU_LD'.)
> So, any particular reason why we would do things differently for
> nvptx?
Let's first work out why binutils assembler vs. linker are handled
differently. ;-)
Grüße
Thomas
-----------------
Siemens Electronic Design Automation GmbH; Anschrift: Arnulfstraße 201, 80634
München; Gesellschaft mit beschränkter Haftung; Geschäftsführer: Thomas
Heurung, Frank Thürauf; Sitz der Gesellschaft: München; Registergericht
München, HRB 106955
