* Can't build GCC, float-int special cases errors
@ 2021-09-13 0:11 André Almeida
2021-09-13 1:05 ` Hongtao Liu
0 siblings, 1 reply; 6+ messages in thread
From: André Almeida @ 2021-09-13 0:11 UTC (permalink / raw)
To: liuhongt; +Cc: italo.niccola, gcc-help
Hi Hongtao Liu,
I was trying to compile GCC, but I'm getting the following error:
during RTL pass: expand
In file included from ../../../gcc/libgcc/soft-fp/soft-fp.h:317,
from ../../../gcc/libgcc/soft-fp/floattihf.c:29:
../../../gcc/libgcc/soft-fp/floattihf.c: In function ‘__floattihf’:
../../../gcc/libgcc/soft-fp/op-1.h:131:13: internal compiler error: in
gen_lowpart_general, at rtlhooks.c:57
131 | (val) = _FP_PACK_RAW_1_flo.flt; \
| ~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~
../../../gcc/libgcc/soft-fp/half.h:84:33: note: in expansion of macro
‘_FP_PACK_RAW_1’
84 | #define FP_PACK_RAW_H(val, X) _FP_PACK_RAW_1 (H, (val), X)
| ^~~~~~~~~~~~~~
../../../gcc/libgcc/soft-fp/floattihf.c:41:3: note: in expansion of
macro ‘FP_PACK_RAW_H’
41 | FP_PACK_RAW_H (a, A);
| ^~~~~~~~~~~~~
I got a similar error on other files as well. I compiled with the
following configuration:
../gcc/configure --prefix=/tmp/gcc --enable-languages=c --disable-bootstrap
I did a git bisect and I found the offending commit that break the build:
57b7c432cce893e1ba60d9b94a9606df6b419379
Revert "Get rid of all float-int special cases in validate_subreg."
I was trying to fix the build myself and upstream the fix, but I didn't
manage to, so I would like to ask for help on how to proceed here.
Thanks!
André
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Can't build GCC, float-int special cases errors
2021-09-13 0:11 Can't build GCC, float-int special cases errors André Almeida
@ 2021-09-13 1:05 ` Hongtao Liu
2021-09-13 23:30 ` André Almeida
2021-09-14 0:36 ` Segher Boessenkool
0 siblings, 2 replies; 6+ messages in thread
From: Hongtao Liu @ 2021-09-13 1:05 UTC (permalink / raw)
To: André Almeida
Cc: liuhongt, italo.niccola, gcc-help, Richard Biener, Segher Boessenkool
Yes, sorry for the inconvenience, the original patch introduced
several regression on power targets, it's decided to revert the patch.
And there's another patch [1] to solve ICE for build HF related libgcc
functions, and the patch is under review.
[1] https://gcc.gnu.org/pipermail/gcc-patches/2021-September/579178.html
On Mon, Sep 13, 2021 at 8:12 AM André Almeida via Gcc-help
<gcc-help@gcc.gnu.org> wrote:
>
> Hi Hongtao Liu,
>
> I was trying to compile GCC, but I'm getting the following error:
>
> during RTL pass: expand
> In file included from ../../../gcc/libgcc/soft-fp/soft-fp.h:317,
> from ../../../gcc/libgcc/soft-fp/floattihf.c:29:
> ../../../gcc/libgcc/soft-fp/floattihf.c: In function ‘__floattihf’:
> ../../../gcc/libgcc/soft-fp/op-1.h:131:13: internal compiler error: in
> gen_lowpart_general, at rtlhooks.c:57
> 131 | (val) = _FP_PACK_RAW_1_flo.flt; \
> | ~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~
> ../../../gcc/libgcc/soft-fp/half.h:84:33: note: in expansion of macro
> ‘_FP_PACK_RAW_1’
> 84 | #define FP_PACK_RAW_H(val, X) _FP_PACK_RAW_1 (H, (val), X)
> | ^~~~~~~~~~~~~~
> ../../../gcc/libgcc/soft-fp/floattihf.c:41:3: note: in expansion of
> macro ‘FP_PACK_RAW_H’
> 41 | FP_PACK_RAW_H (a, A);
> | ^~~~~~~~~~~~~
>
> I got a similar error on other files as well. I compiled with the
> following configuration:
>
> ../gcc/configure --prefix=/tmp/gcc --enable-languages=c --disable-bootstrap
>
> I did a git bisect and I found the offending commit that break the build:
>
> 57b7c432cce893e1ba60d9b94a9606df6b419379
>
> Revert "Get rid of all float-int special cases in validate_subreg."
>
> I was trying to fix the build myself and upstream the fix, but I didn't
> manage to, so I would like to ask for help on how to proceed here.
>
> Thanks!
> André
--
BR,
Hongtao
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Can't build GCC, float-int special cases errors
2021-09-13 1:05 ` Hongtao Liu
@ 2021-09-13 23:30 ` André Almeida
2021-09-14 1:24 ` Hongtao Liu
2021-09-14 0:36 ` Segher Boessenkool
1 sibling, 1 reply; 6+ messages in thread
From: André Almeida @ 2021-09-13 23:30 UTC (permalink / raw)
To: Hongtao Liu
Cc: liuhongt, italo.niccola, gcc-help, Richard Biener, Segher Boessenkool
I confirm that this patch fix the issue!
I found other build breaks in some older commits, do you have any tips
to on how to do git bisect on gcc repo?
Às 22:05 de 12/09/21, Hongtao Liu escreveu:
> Yes, sorry for the inconvenience, the original patch introduced
> several regression on power targets, it's decided to revert the patch.
> And there's another patch [1] to solve ICE for build HF related libgcc
> functions, and the patch is under review.
> [1] https://gcc.gnu.org/pipermail/gcc-patches/2021-September/579178.html
>
> On Mon, Sep 13, 2021 at 8:12 AM André Almeida via Gcc-help
> <gcc-help@gcc.gnu.org> wrote:
>>
>> Hi Hongtao Liu,
>>
>> I was trying to compile GCC, but I'm getting the following error:
>>
>> during RTL pass: expand
>> In file included from ../../../gcc/libgcc/soft-fp/soft-fp.h:317,
>> from ../../../gcc/libgcc/soft-fp/floattihf.c:29:
>> ../../../gcc/libgcc/soft-fp/floattihf.c: In function ‘__floattihf’:
>> ../../../gcc/libgcc/soft-fp/op-1.h:131:13: internal compiler error: in
>> gen_lowpart_general, at rtlhooks.c:57
>> 131 | (val) = _FP_PACK_RAW_1_flo.flt; \
>> | ~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~
>> ../../../gcc/libgcc/soft-fp/half.h:84:33: note: in expansion of macro
>> ‘_FP_PACK_RAW_1’
>> 84 | #define FP_PACK_RAW_H(val, X) _FP_PACK_RAW_1 (H, (val), X)
>> | ^~~~~~~~~~~~~~
>> ../../../gcc/libgcc/soft-fp/floattihf.c:41:3: note: in expansion of
>> macro ‘FP_PACK_RAW_H’
>> 41 | FP_PACK_RAW_H (a, A);
>> | ^~~~~~~~~~~~~
>>
>> I got a similar error on other files as well. I compiled with the
>> following configuration:
>>
>> ../gcc/configure --prefix=/tmp/gcc --enable-languages=c --disable-bootstrap
>>
>> I did a git bisect and I found the offending commit that break the build:
>>
>> 57b7c432cce893e1ba60d9b94a9606df6b419379
>>
>> Revert "Get rid of all float-int special cases in validate_subreg."
>>
>> I was trying to fix the build myself and upstream the fix, but I didn't
>> manage to, so I would like to ask for help on how to proceed here.
>>
>> Thanks!
>> André
>
>
>
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Can't build GCC, float-int special cases errors
2021-09-13 1:05 ` Hongtao Liu
2021-09-13 23:30 ` André Almeida
@ 2021-09-14 0:36 ` Segher Boessenkool
1 sibling, 0 replies; 6+ messages in thread
From: Segher Boessenkool @ 2021-09-14 0:36 UTC (permalink / raw)
To: Hongtao Liu
Cc: André Almeida, liuhongt, italo.niccola, gcc-help, Richard Biener
On Mon, Sep 13, 2021 at 09:05:26AM +0800, Hongtao Liu wrote:
> Yes, sorry for the inconvenience, the original patch introduced
> several regression on power targets, it's decided to revert the patch.
Not just on rs6000 (as your messages for the reversion said, but not the
patch you committed).
> And there's another patch [1] to solve ICE for build HF related libgcc
> functions, and the patch is under review.
> [1] https://gcc.gnu.org/pipermail/gcc-patches/2021-September/579178.html
That patch mail does not say it solves an ICE during bootstrap. It can
help people prioritise things if you mention that :-)
Segher
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Can't build GCC, float-int special cases errors
2021-09-13 23:30 ` André Almeida
@ 2021-09-14 1:24 ` Hongtao Liu
2021-09-14 5:01 ` Hongtao Liu
0 siblings, 1 reply; 6+ messages in thread
From: Hongtao Liu @ 2021-09-14 1:24 UTC (permalink / raw)
To: André Almeida
Cc: liuhongt, italo.niccola, gcc-help, Richard Biener, Segher Boessenkool
On Tue, Sep 14, 2021 at 8:07 AM André Almeida <andrealmeid@riseup.net> wrote:
>
> I confirm that this patch fix the issue!
>
> I found other build breaks in some older commits, do you have any tips
> to on how to do git bisect on gcc repo?
You can create a branch based on master, and rebase my fix patch to
make it just after the revert commits, then you can do regular gcc
bisect for guilty commit.
>
> Às 22:05 de 12/09/21, Hongtao Liu escreveu:
> > Yes, sorry for the inconvenience, the original patch introduced
> > several regression on power targets, it's decided to revert the patch.
> > And there's another patch [1] to solve ICE for build HF related libgcc
> > functions, and the patch is under review.
> > [1] https://gcc.gnu.org/pipermail/gcc-patches/2021-September/579178.html
> >
> > On Mon, Sep 13, 2021 at 8:12 AM André Almeida via Gcc-help
> > <gcc-help@gcc.gnu.org> wrote:
> >>
> >> Hi Hongtao Liu,
> >>
> >> I was trying to compile GCC, but I'm getting the following error:
> >>
> >> during RTL pass: expand
> >> In file included from ../../../gcc/libgcc/soft-fp/soft-fp.h:317,
> >> from ../../../gcc/libgcc/soft-fp/floattihf.c:29:
> >> ../../../gcc/libgcc/soft-fp/floattihf.c: In function ‘__floattihf’:
> >> ../../../gcc/libgcc/soft-fp/op-1.h:131:13: internal compiler error: in
> >> gen_lowpart_general, at rtlhooks.c:57
> >> 131 | (val) = _FP_PACK_RAW_1_flo.flt; \
> >> | ~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~
> >> ../../../gcc/libgcc/soft-fp/half.h:84:33: note: in expansion of macro
> >> ‘_FP_PACK_RAW_1’
> >> 84 | #define FP_PACK_RAW_H(val, X) _FP_PACK_RAW_1 (H, (val), X)
> >> | ^~~~~~~~~~~~~~
> >> ../../../gcc/libgcc/soft-fp/floattihf.c:41:3: note: in expansion of
> >> macro ‘FP_PACK_RAW_H’
> >> 41 | FP_PACK_RAW_H (a, A);
> >> | ^~~~~~~~~~~~~
> >>
> >> I got a similar error on other files as well. I compiled with the
> >> following configuration:
> >>
> >> ../gcc/configure --prefix=/tmp/gcc --enable-languages=c --disable-bootstrap
> >>
> >> I did a git bisect and I found the offending commit that break the build:
> >>
> >> 57b7c432cce893e1ba60d9b94a9606df6b419379
> >>
> >> Revert "Get rid of all float-int special cases in validate_subreg."
> >>
> >> I was trying to fix the build myself and upstream the fix, but I didn't
> >> manage to, so I would like to ask for help on how to proceed here.
> >>
> >> Thanks!
> >> André
> >
> >
> >
--
BR,
Hongtao
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Can't build GCC, float-int special cases errors
2021-09-14 1:24 ` Hongtao Liu
@ 2021-09-14 5:01 ` Hongtao Liu
0 siblings, 0 replies; 6+ messages in thread
From: Hongtao Liu @ 2021-09-14 5:01 UTC (permalink / raw)
To: André Almeida
Cc: liuhongt, italo.niccola, gcc-help, Richard Biener, Segher Boessenkool
On Tue, Sep 14, 2021 at 9:24 AM Hongtao Liu <crazylht@gmail.com> wrote:
>
> On Tue, Sep 14, 2021 at 8:07 AM André Almeida <andrealmeid@riseup.net> wrote:
> >
> > I confirm that this patch fix the issue!
> >
> > I found other build breaks in some older commits, do you have any tips
> > to on how to do git bisect on gcc repo?
> You can create a branch based on master, and rebase my fix patch to
> make it just after the revert commits, then you can do regular gcc
> bisect for guilty commit.
i've compare gcc testsuite between r12-3456 and r12-3484, there're 5
new failure testcase
unix/-m32: c-c++-common/auto-init-padding-1.c -Wc++-compat
scan-tree-dump gimple ".DEFERRED_INIT \\(24, 1, 0\\)"
unix/-m32: c-c++-common/auto-init-padding-1.c -std=gnu++14
scan-tree-dump gimple ".DEFERRED_INIT \\(24, 1, 0\\)"
unix/-m32: c-c++-common/auto-init-padding-1.c -std=gnu++17
scan-tree-dump gimple ".DEFERRED_INIT \\(24, 1, 0\\)"
unix/-m32: c-c++-common/auto-init-padding-1.c -std=gnu++2a
scan-tree-dump gimple ".DEFERRED_INIT \\(24, 1, 0\\)"
unix/-m32: c-c++-common/auto-init-padding-1.c -std=gnu++98
scan-tree-dump gimple ".DEFERRED_INIT \\(24, 1, 0\\)"
And they seem to be fix in the lastest trunk.
> >
> > Às 22:05 de 12/09/21, Hongtao Liu escreveu:
> > > Yes, sorry for the inconvenience, the original patch introduced
> > > several regression on power targets, it's decided to revert the patch.
> > > And there's another patch [1] to solve ICE for build HF related libgcc
> > > functions, and the patch is under review.
> > > [1] https://gcc.gnu.org/pipermail/gcc-patches/2021-September/579178.html
> > >
> > > On Mon, Sep 13, 2021 at 8:12 AM André Almeida via Gcc-help
> > > <gcc-help@gcc.gnu.org> wrote:
> > >>
> > >> Hi Hongtao Liu,
> > >>
> > >> I was trying to compile GCC, but I'm getting the following error:
> > >>
> > >> during RTL pass: expand
> > >> In file included from ../../../gcc/libgcc/soft-fp/soft-fp.h:317,
> > >> from ../../../gcc/libgcc/soft-fp/floattihf.c:29:
> > >> ../../../gcc/libgcc/soft-fp/floattihf.c: In function ‘__floattihf’:
> > >> ../../../gcc/libgcc/soft-fp/op-1.h:131:13: internal compiler error: in
> > >> gen_lowpart_general, at rtlhooks.c:57
> > >> 131 | (val) = _FP_PACK_RAW_1_flo.flt; \
> > >> | ~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~
> > >> ../../../gcc/libgcc/soft-fp/half.h:84:33: note: in expansion of macro
> > >> ‘_FP_PACK_RAW_1’
> > >> 84 | #define FP_PACK_RAW_H(val, X) _FP_PACK_RAW_1 (H, (val), X)
> > >> | ^~~~~~~~~~~~~~
> > >> ../../../gcc/libgcc/soft-fp/floattihf.c:41:3: note: in expansion of
> > >> macro ‘FP_PACK_RAW_H’
> > >> 41 | FP_PACK_RAW_H (a, A);
> > >> | ^~~~~~~~~~~~~
> > >>
> > >> I got a similar error on other files as well. I compiled with the
> > >> following configuration:
> > >>
> > >> ../gcc/configure --prefix=/tmp/gcc --enable-languages=c --disable-bootstrap
> > >>
> > >> I did a git bisect and I found the offending commit that break the build:
> > >>
> > >> 57b7c432cce893e1ba60d9b94a9606df6b419379
> > >>
> > >> Revert "Get rid of all float-int special cases in validate_subreg."
> > >>
> > >> I was trying to fix the build myself and upstream the fix, but I didn't
> > >> manage to, so I would like to ask for help on how to proceed here.
> > >>
> > >> Thanks!
> > >> André
> > >
> > >
> > >
>
>
>
> --
> BR,
> Hongtao
--
BR,
Hongtao
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2021-09-14 5:02 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-09-13 0:11 Can't build GCC, float-int special cases errors André Almeida
2021-09-13 1:05 ` Hongtao Liu
2021-09-13 23:30 ` André Almeida
2021-09-14 1:24 ` Hongtao Liu
2021-09-14 5:01 ` Hongtao Liu
2021-09-14 0:36 ` Segher Boessenkool
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).