GCC 9.5 Release Candidate available

GCC 9.5 Release Candidate available

[Richard Biener]

The first release candidate for GCC 9.5 is available from

https://sourceware.org/pub/gcc/snapshots/9.5.0-RC-20220520/

and shortly its mirrors.  It has been generated from git commit
1bc79c506205b6a5db82897340bdebaaf7ada934.

I have so far bootstrapped and tested the release candidate
on x86_64-suse-linux.

Please test it and report any issues to bugzilla.

If all goes well I'd like to release 9.5 on Friday, May 27th,
which will then close the branch.

Re: GCC 9.5 Release Candidate available

[Iain Sandoe]

Hi

> On 20 May 2022, at 09:02, Richard Biener via Gcc <gcc@gcc.gnu.org> wrote:

> The first release candidate for GCC 9.5 is available from
> 
> https://sourceware.org/pub/gcc/snapshots/9.5.0-RC-20220520/
> 
> and shortly its mirrors.  It has been generated from git commit
> 1bc79c506205b6a5db82897340bdebaaf7ada934.
> 
> I have so far bootstrapped and tested the release candidate
> on x86_64-suse-linux.

I have bootstrapped (using GCC5.4 on darwin9 and GCC7.5 elsewhere) r9-10192 on:
i686-darwin9,17
powerpc-darwin9
x86_64-darwin10 to 21.

As, expected (since I was not able to find enough time to do the backports),
although bootstrap succeeds on darwin21 (macOS 12) the resulting compiler
is not really usable.  I will have to provide a darwin branch with the necessary
changes.

One observation outside of this:

Several of the testsuite runs hung with cc1 spinning in reload for pr88414.
I was not really able to correlate exactly with CPU / tuning chosen.

Pretty sure this is a regression - I do not recall the testsuite hanging (for anything
other than D) for years.

thanks
Iain

Re: GCC 9.5 Release Candidate available

[Richard Biener]

On Sun, 22 May 2022, Iain Sandoe wrote:

> Hi
> 
> > On 20 May 2022, at 09:02, Richard Biener via Gcc <gcc@gcc.gnu.org> wrote:
> 
> > The first release candidate for GCC 9.5 is available from
> > 
> > https://sourceware.org/pub/gcc/snapshots/9.5.0-RC-20220520/
> > 
> > and shortly its mirrors.  It has been generated from git commit
> > 1bc79c506205b6a5db82897340bdebaaf7ada934.
> > 
> > I have so far bootstrapped and tested the release candidate
> > on x86_64-suse-linux.
> 
> I have bootstrapped (using GCC5.4 on darwin9 and GCC7.5 elsewhere) r9-10192 on:
> i686-darwin9,17
> powerpc-darwin9
> x86_64-darwin10 to 21.
> 
> As, expected (since I was not able to find enough time to do the backports),
> although bootstrap succeeds on darwin21 (macOS 12) the resulting compiler
> is not really usable.  I will have to provide a darwin branch with the necessary
> changes.
> 
> One observation outside of this:
> 
> Several of the testsuite runs hung with cc1 spinning in reload for pr88414.
> I was not really able to correlate exactly with CPU / tuning chosen.
> 
> Pretty sure this is a regression - I do not recall the testsuite hanging (for anything
> other than D) for years.

That's gcc.target/i386/pr88414.c?

Can you be more specific on the target the issue occurs on (so one
can maybe try with a cross?).  Bisecting would be most helpful of
course, if it's some of the recent backports reversion would be
most appropriate at this point.

Thanks,
Richard.

Re: GCC 9.5 Release Candidate available

[Iain Sandoe]

Hi Richard,

> On 23 May 2022, at 07:27, Richard Biener <rguenther@suse.de> wrote:
> 
> On Sun, 22 May 2022, Iain Sandoe wrote:
> 
>> Hi
>> 
>>> On 20 May 2022, at 09:02, Richard Biener via Gcc <gcc@gcc.gnu.org> wrote:
>> 
>>> The first release candidate for GCC 9.5 is available from
>>> 
>>> https://sourceware.org/pub/gcc/snapshots/9.5.0-RC-20220520/
>>> 
>>> and shortly its mirrors.  It has been generated from git commit
>>> 1bc79c506205b6a5db82897340bdebaaf7ada934.
>>> 
>>> I have so far bootstrapped and tested the release candidate
>>> on x86_64-suse-linux.
>> 
>> I have bootstrapped (using GCC5.4 on darwin9 and GCC7.5 elsewhere) r9-10192 on:
>> i686-darwin9,17
>> powerpc-darwin9
>> x86_64-darwin10 to 21.
>> 
>> As, expected (since I was not able to find enough time to do the backports),
>> although bootstrap succeeds on darwin21 (macOS 12) the resulting compiler
>> is not really usable.  I will have to provide a darwin branch with the necessary
>> changes.
>> 
>> One observation outside of this:
>> 
>> Several of the testsuite runs hung with cc1 spinning in reload for pr88414.
>> I was not really able to correlate exactly with CPU / tuning chosen.
>> 
>> Pretty sure this is a regression - I do not recall the testsuite hanging (for anything
>> other than D) for years.
> 
> That's gcc.target/i386/pr88414.c?
> 
> Can you be more specific on the target the issue occurs on (so one
> can maybe try with a cross?).  Bisecting would be most helpful of
> course, if it's some of the recent backports reversion would be
> most appropriate at this point.

several cases, here’s the most modern with an m32 multilib:

Configured with: /src-local/gcc-git-9/configure --prefix=/opt/iains/x86_64-apple-darwin17/gcc-9-wip --build=x86_64-apple-darwin17 --with-sysroot=/Library//Developer/CommandLineTools/SDKs/MacOSX.sdk --with-as=/XC/9.4/usr/bin/as --with-ld=/XC/9.4/usr/bin/ld --enable-languages=all CC=x86_64-apple-darwin17-gcc CXX=x86_64-apple-darwin17-g++

log info:

[ note the m64 multilib, also emits the diagnostic, but it does not appear to spin. ]

/scratch/10-13-his/gcc-9-wip/gcc/xgcc -B/scratch/10-13-his/gcc-9-wip/gcc/ /src-local/gcc-git-9/gcc/testsuite/gcc.target/i386/pr88414.c -m32 -fno-diagnostics-show-caret -fno-diagnostics-show-line-numbers -fdiagnostics-color=never -O1 -ftrapv -S -o pr88414.s
/src-local/gcc-git-9/gcc/testsuite/gcc.target/i386/pr88414.c: In function 'foo':
/src-local/gcc-git-9/gcc/testsuite/gcc.target/i386/pr88414.c:15:7: error: 'asm' operand has impossible constraints
 got a INT signal, interrupted by user // (manually killed it)

Iain

Re: GCC 9.5 Release Candidate available

[Iain Sandoe]

> On 23 May 2022, at 07:50, Iain Sandoe <iain@sandoe.co.uk> wrote:
> 
> Hi Richard,
> 
>> On 23 May 2022, at 07:27, Richard Biener <rguenther@suse.de> wrote:
>> 
>> On Sun, 22 May 2022, Iain Sandoe wrote:
>> 
>>> Hi
>>> 
>>>> On 20 May 2022, at 09:02, Richard Biener via Gcc <gcc@gcc.gnu.org> wrote:
>>> 
>>>> The first release candidate for GCC 9.5 is available from
>>>> 
>>>> https://sourceware.org/pub/gcc/snapshots/9.5.0-RC-20220520/
>>>> 
>>>> and shortly its mirrors.  It has been generated from git commit
>>>> 1bc79c506205b6a5db82897340bdebaaf7ada934.
>>>> 
>>>> I have so far bootstrapped and tested the release candidate
>>>> on x86_64-suse-linux.
>>> 
>>> I have bootstrapped (using GCC5.4 on darwin9 and GCC7.5 elsewhere) r9-10192 on:
>>> i686-darwin9,17
>>> powerpc-darwin9
>>> x86_64-darwin10 to 21.
>>> 
>>> As, expected (since I was not able to find enough time to do the backports),
>>> although bootstrap succeeds on darwin21 (macOS 12) the resulting compiler
>>> is not really usable.  I will have to provide a darwin branch with the necessary
>>> changes.
>>> 
>>> One observation outside of this:
>>> 
>>> Several of the testsuite runs hung with cc1 spinning in reload for pr88414.
>>> I was not really able to correlate exactly with CPU / tuning chosen.
>>> 
>>> Pretty sure this is a regression - I do not recall the testsuite hanging (for anything
>>> other than D) for years.
>> 
>> That's gcc.target/i386/pr88414.c?
>> 
>> Can you be more specific on the target the issue occurs on (so one
>> can maybe try with a cross?).  Bisecting would be most helpful of
>> course, if it's some of the recent backports reversion would be
>> most appropriate at this point.
> 
> several cases, here’s the most modern with an m32 multilib:
> 
> Configured with: /src-local/gcc-git-9/configure --prefix=/opt/iains/x86_64-apple-darwin17/gcc-9-wip --build=x86_64-apple-darwin17 --with-sysroot=/Library//Developer/CommandLineTools/SDKs/MacOSX.sdk --with-as=/XC/9.4/usr/bin/as --with-ld=/XC/9.4/usr/bin/ld --enable-languages=all CC=x86_64-apple-darwin17-gcc CXX=x86_64-apple-darwin17-g++
> 
> log info:
> 
> [ note the m64 multilib, also emits the diagnostic, but it does not appear to spin. ]
> 
> /scratch/10-13-his/gcc-9-wip/gcc/xgcc -B/scratch/10-13-his/gcc-9-wip/gcc/ /src-local/gcc-git-9/gcc/testsuite/gcc.target/i386/pr88414.c -m32 -fno-diagnostics-show-caret -fno-diagnostics-show-line-numbers -fdiagnostics-color=never -O1 -ftrapv -S -o pr88414.s
> /src-local/gcc-git-9/gcc/testsuite/gcc.target/i386/pr88414.c: In function 'foo':
> /src-local/gcc-git-9/gcc/testsuite/gcc.target/i386/pr88414.c:15:7: error: 'asm' operand has impossible constraints
> got a INT signal, interrupted by user // (manually killed it)

Hmm, this is not a regression - (although the problem is real) - I also see it on 9.4 and 9.3 at least, 
apologies for the noise.
Iain

Re: GCC 9.5 Release Candidate available

[William Seurer]

On 5/20/22 3:02 AM, Richard Biener via Gcc wrote:
> The first release candidate for GCC 9.5 is available from
>
> https://sourceware.org/pub/gcc/snapshots/9.5.0-RC-20220520/
>
> and shortly its mirrors.  It has been generated from git commit
> 1bc79c506205b6a5db82897340bdebaaf7ada934.
>
> I have so far bootstrapped and tested the release candidate
> on x86_64-suse-linux.
>
> Please test it and report any issues to bugzilla.
>
> If all goes well I'd like to release 9.5 on Friday, May 27th,
> which will then close the branch.


I tried the RC on powerpc64 power 7 BE, power 8 BE and LE, and power 9 
and 10 LE and saw nothing unexpected.

Re: GCC 9.5 Release Candidate available

[Andrew Pinski]

On Fri, May 20, 2022 at 1:03 AM Richard Biener via Gcc <gcc@gcc.gnu.org> wrote:
>
>
> The first release candidate for GCC 9.5 is available from
>
> https://sourceware.org/pub/gcc/snapshots/9.5.0-RC-20220520/
>
> and shortly its mirrors.  It has been generated from git commit
> 1bc79c506205b6a5db82897340bdebaaf7ada934.
>
> I have so far bootstrapped and tested the release candidate
> on x86_64-suse-linux.
>
> Please test it and report any issues to bugzilla.
>
> If all goes well I'd like to release 9.5 on Friday, May 27th,
> which will then close the branch.

A brand new regression was just filed:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105732

This is a regression from 9.4.0 even.

Thanks,
Andrew