Re: [r14-5930 Regression] FAIL: gcc.c-torture/compile/libcall-2.c -Os (test for excess errors) on Linux/x86_64

Re: [r14-5930 Regression] FAIL: gcc.c-torture/compile/libcall-2.c -Os (test for excess errors) on Linux/x86_64

[FX Coudert]

That commit makes gcc.target/i386/libcall-1.c on darwin:

FAIL: gcc.target/i386/libcall-1.c scan-assembler globl\t__divti3

because the pattern is not found, the only mention of divti3 in the generated assembly is:

LCFI0:
        movabsq $_b@GOTOFF, %rdx
        movabsq $___divti3@PLTOFF, %rax
        leaq    L2(%rip), %r15
        pushq   %rbx


The source code is:

---
/* Make sure that external refences for libcalls are generated even for
   indirect calls.  */

/* { dg-do compile { target int128 } } */
/* { dg-options "-O2 -mcmodel=large" } */
/* { dg-final { scan-assembler "globl\t__divti3" } } */

__int128 a, b; void foo () { a = a / b; }
---

Looking at, for example, gcc.target/i386/falign-functions-3.c it seems that test avoids scanning for global references on darwin. Probably the new test needs the same exception.

FX

Re: [r14-5930 Regression] FAIL: gcc.c-torture/compile/libcall-2.c -Os (test for excess errors) on Linux/x86_64

[Iain Sandoe]

Hi FX,

> On 1 Dec 2023, at 13:55, FX Coudert <fxcoudert@gmail.com> wrote:
> 
> That commit makes gcc.target/i386/libcall-1.c on darwin:
> 
> FAIL: gcc.target/i386/libcall-1.c scan-assembler globl\t__divti3
> 
> because the pattern is not found, the only mention of divti3 in the generated assembly is:
> 
> LCFI0:
>        movabsq $_b@GOTOFF, %rdx
>        movabsq $___divti3@PLTOFF, %rax
>        leaq    L2(%rip), %r15
>        pushq   %rbx
> 
> 
> The source code is:
> 
> ---
> /* Make sure that external refences for libcalls are generated even for
>   indirect calls.  */
> 
> /* { dg-do compile { target int128 } } */
> /* { dg-options "-O2 -mcmodel=large" } */

mcmodel=large s not supported (yet) on any Darwin arch [PR90698], so the test needs skipping or xfailing, I think (either way with a reference to the PR).

Iain

> /* { dg-final { scan-assembler "globl\t__divti3" } } */
> 
> __int128 a, b; void foo () { a = a / b; }
> ---
> 
> Looking at, for example, gcc.target/i386/falign-functions-3.c it seems that test avoids scanning for global references on darwin. Probably the new test needs the same exception.
> 
> FX

Re: [r14-5930 Regression] FAIL: gcc.c-torture/compile/libcall-2.c -Os (test for excess errors) on Linux/x86_64

[FX Coudert]

> mcmodel=large s not supported (yet) on any Darwin arch [PR90698], so the test needs skipping or xfailing, I think (either way with a reference to the PR).

Pushed as https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=b74981b5cf32ebf4bfffd25e7174b5c80243447a

FX

Re: [r14-5930 Regression] FAIL: gcc.c-torture/compile/libcall-2.c -Os (test for excess errors) on Linux/x86_64

[Jose E. Marchesi]

>> mcmodel=large s not supported (yet) on any Darwin arch [PR90698], so
> the test needs skipping or xfailing, I think (either way with a
> reference to the PR).
>
> Pushed as https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=b74981b5cf32ebf4bfffd25e7174b5c80243447a

Thanks for fixing this.

Re: [r14-5930 Regression] FAIL: gcc.c-torture/compile/libcall-2.c -Os (test for excess errors) on Linux/x86_64

[FX Coudert]

>> Pushed as https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=b74981b5cf32ebf4bfffd25e7174b5c80243447a

Somehow I pushed the wrong commit, we should skip the test and not xfail.
This showed up in https://gcc.gnu.org/pipermail/gcc-testresults/2023-December/802839.html

So, new commit push as obvious fixup: https://gcc.gnu.org/git/?p=gcc.git;a=commit;h=53e954a673a0d6ac80ab1f0591ea4f751e67374c

FX

Re: [r14-5930 Regression] FAIL: gcc.c-torture/compile/libcall-2.c -Os (test for excess errors) on Linux/x86_64

[Jose E. Marchesi]

> On Wed, Nov 29, 2023 at 09:20:55AM +0100, Jose E. Marchesi wrote:
>> OK to the patch below?
>> 
>> >From a697f9e9442d96d3eda228b825dd8e8e06206c03 Mon Sep 17 00:00:00 2001
>> From: "Jose E. Marchesi" <jose.marchesi@oracle.com>
>> Date: Wed, 29 Nov 2023 09:16:48 +0100
>> Subject: [PATCH] testsuite: move gcc.c-torture/compile/libcall-2.c to
>>  gcc.target/i386/libcall-1.c
>> 
>> This patch relocates a test that is really x86 specific, and changes
>> it to use check_effective_target_int128.
>> 
>> gcc/testsuite/ChangeLog
>> 
>> 	* gcc.c-torture/compile/libcall-2.c: Remove.
>> 	* gcc.target/i386/libcall-1.c: Moved from
>> 	gcc.c-torture/compile/libcall-2.c and adapted to use
>> 	effective-target for int128_t.
>
> Yes.

Thanks.  Pushed.

>
>> diff --git a/gcc/testsuite/gcc.c-torture/compile/libcall-2.c b/gcc/testsuite/gcc.target/i386/libcall-1.c
>> similarity index 50%
>> rename from gcc/testsuite/gcc.c-torture/compile/libcall-2.c
>> rename to gcc/testsuite/gcc.target/i386/libcall-1.c
>> index 9b889172025..8bd6684a925 100644
>> --- a/gcc/testsuite/gcc.c-torture/compile/libcall-2.c
>> +++ b/gcc/testsuite/gcc.target/i386/libcall-1.c
>> @@ -1,10 +1,8 @@
>>  /* Make sure that external refences for libcalls are generated even for
>>     indirect calls.  */
>>  
>> -/* { dg-do compile } */
>> -/* __int128 is not supported in x86 -m32.  */
>> -/* { dg-skip-if "" { ! { x86_64-*-* && { ! ilp32 } } } } */
>> -/* { dg-options "-O2 -mcmodel=large" { target x86_64-*-* } } */
>> +/* { dg-do compile { target int128 } } */
>> +/* { dg-options "-O2 -mcmodel=large" } */
>>  /* { dg-final { scan-assembler "globl\t__divti3" } } */
>>  
>>  __int128 a, b; void foo () { a = a / b; }
>> -- 
>> 2.30.2
>
> 	Jakub

Re: [r14-5930 Regression] FAIL: gcc.c-torture/compile/libcall-2.c -Os (test for excess errors) on Linux/x86_64

[Jakub Jelinek]

On Wed, Nov 29, 2023 at 09:20:55AM +0100, Jose E. Marchesi wrote:
> OK to the patch below?
> 
> >From a697f9e9442d96d3eda228b825dd8e8e06206c03 Mon Sep 17 00:00:00 2001
> From: "Jose E. Marchesi" <jose.marchesi@oracle.com>
> Date: Wed, 29 Nov 2023 09:16:48 +0100
> Subject: [PATCH] testsuite: move gcc.c-torture/compile/libcall-2.c to
>  gcc.target/i386/libcall-1.c
> 
> This patch relocates a test that is really x86 specific, and changes
> it to use check_effective_target_int128.
> 
> gcc/testsuite/ChangeLog
> 
> 	* gcc.c-torture/compile/libcall-2.c: Remove.
> 	* gcc.target/i386/libcall-1.c: Moved from
> 	gcc.c-torture/compile/libcall-2.c and adapted to use
> 	effective-target for int128_t.

Yes.

> diff --git a/gcc/testsuite/gcc.c-torture/compile/libcall-2.c b/gcc/testsuite/gcc.target/i386/libcall-1.c
> similarity index 50%
> rename from gcc/testsuite/gcc.c-torture/compile/libcall-2.c
> rename to gcc/testsuite/gcc.target/i386/libcall-1.c
> index 9b889172025..8bd6684a925 100644
> --- a/gcc/testsuite/gcc.c-torture/compile/libcall-2.c
> +++ b/gcc/testsuite/gcc.target/i386/libcall-1.c
> @@ -1,10 +1,8 @@
>  /* Make sure that external refences for libcalls are generated even for
>     indirect calls.  */
>  
> -/* { dg-do compile } */
> -/* __int128 is not supported in x86 -m32.  */
> -/* { dg-skip-if "" { ! { x86_64-*-* && { ! ilp32 } } } } */
> -/* { dg-options "-O2 -mcmodel=large" { target x86_64-*-* } } */
> +/* { dg-do compile { target int128 } } */
> +/* { dg-options "-O2 -mcmodel=large" } */
>  /* { dg-final { scan-assembler "globl\t__divti3" } } */
>  
>  __int128 a, b; void foo () { a = a / b; }
> -- 
> 2.30.2

	Jakub

Re: [r14-5930 Regression] FAIL: gcc.c-torture/compile/libcall-2.c -Os (test for excess errors) on Linux/x86_64

[Jose E. Marchesi]

> On Wed, Nov 29, 2023 at 07:51:15AM +0100, Jose E. Marchesi wrote:
>> > FAIL: gcc.c-torture/compile/libcall-2.c   -O1  (test for excess errors)
>> > FAIL: gcc.c-torture/compile/libcall-2.c -O2 -flto
>> > -fno-use-linker-plugin -flto-partition=none (test for excess errors)
>> > FAIL: gcc.c-torture/compile/libcall-2.c -O2 -flto -fuse-linker-plugin
>> > -fno-fat-lto-objects (test for excess errors)
>> > FAIL: gcc.c-torture/compile/libcall-2.c   -O2  (test for excess errors)
>> > FAIL: gcc.c-torture/compile/libcall-2.c   -O3 -g  (test for excess errors)
>> > FAIL: gcc.c-torture/compile/libcall-2.c   -Os  (test for excess errors)
>> 
>> Sorry about the regression.
>> 
>> I installed the patch below to skip the test if the target is not x86_64
>> in lp64, as obvious.  This should fix the issue.
>> 
>> >From 4ed0740c6e807460ce79a351094329fdeb551545 Mon Sep 17 00:00:00 2001
>> From: "Jose E. Marchesi" <jose.marchesi@oracle.com>
>> Date: Wed, 29 Nov 2023 07:44:59 +0100
>> Subject: [PATCH] testsuite: fix gcc.c-torture/compile/libcall-2.c in -m32
>> 
>> This test relies on having __int128 in x86_64 targets, which is only
>> available in -m64.
>> 
>> gcc/testsuite/ChangeLog
>> 
>> 	* gcc.c-torture/compile/libcall-2.c: Skip test in -m32.
>> ---
>>  gcc/testsuite/gcc.c-torture/compile/libcall-2.c | 2 ++
>>  1 file changed, 2 insertions(+)
>> 
>> diff --git a/gcc/testsuite/gcc.c-torture/compile/libcall-2.c b/gcc/testsuite/gcc.c-torture/compile/libcall-2.c
>> index b33944c83ff..9b889172025 100644
>> --- a/gcc/testsuite/gcc.c-torture/compile/libcall-2.c
>> +++ b/gcc/testsuite/gcc.c-torture/compile/libcall-2.c
>> @@ -2,6 +2,8 @@
>>     indirect calls.  */
>>  
>>  /* { dg-do compile } */
>> +/* __int128 is not supported in x86 -m32.  */
>> +/* { dg-skip-if "" { ! { x86_64-*-* && { ! ilp32 } } } } */
>>  /* { dg-options "-O2 -mcmodel=large" { target x86_64-*-* } } */
>>  /* { dg-final { scan-assembler "globl\t__divti3" } } */
>
> This is not correct.
> When a test uses __int128, it should be guarded with int128 effective
> target.
> But, as the test doesn't really test anything on non-x86 nor ia32,
> it doesn't belong to gcc.c-torture/compile/ at all, it is x86 specific
> test which should be moved to gcc.target/i386/libcall-1.c
> And, should have
> /* { dg-do compile { target int128 } } */
> /* { dg-options "-O2 -mcmodel=large" } */
> /* { dg-final { scan-assembler "globl\t__divti3" } } */
> I guess no need to bother with the extra guard for -mcmodel=large,
> because -m32/-mx32 don't have __int128 support, and x86_64-*-*
> is incorrect anyway (because with that target one can have all
> of -m32/-m64/-mx32).

OK to the patch below?

From a697f9e9442d96d3eda228b825dd8e8e06206c03 Mon Sep 17 00:00:00 2001
From: "Jose E. Marchesi" <jose.marchesi@oracle.com>
Date: Wed, 29 Nov 2023 09:16:48 +0100
Subject: [PATCH] testsuite: move gcc.c-torture/compile/libcall-2.c to
 gcc.target/i386/libcall-1.c

This patch relocates a test that is really x86 specific, and changes
it to use check_effective_target_int128.

gcc/testsuite/ChangeLog

	* gcc.c-torture/compile/libcall-2.c: Remove.
	* gcc.target/i386/libcall-1.c: Moved from
	gcc.c-torture/compile/libcall-2.c and adapted to use
	effective-target for int128_t.
---
 .../compile/libcall-2.c => gcc.target/i386/libcall-1.c}     | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
 rename gcc/testsuite/{gcc.c-torture/compile/libcall-2.c => gcc.target/i386/libcall-1.c} (50%)

diff --git a/gcc/testsuite/gcc.c-torture/compile/libcall-2.c b/gcc/testsuite/gcc.target/i386/libcall-1.c
similarity index 50%
rename from gcc/testsuite/gcc.c-torture/compile/libcall-2.c
rename to gcc/testsuite/gcc.target/i386/libcall-1.c
index 9b889172025..8bd6684a925 100644
--- a/gcc/testsuite/gcc.c-torture/compile/libcall-2.c
+++ b/gcc/testsuite/gcc.target/i386/libcall-1.c
@@ -1,10 +1,8 @@
 /* Make sure that external refences for libcalls are generated even for
    indirect calls.  */
 
-/* { dg-do compile } */
-/* __int128 is not supported in x86 -m32.  */
-/* { dg-skip-if "" { ! { x86_64-*-* && { ! ilp32 } } } } */
-/* { dg-options "-O2 -mcmodel=large" { target x86_64-*-* } } */
+/* { dg-do compile { target int128 } } */
+/* { dg-options "-O2 -mcmodel=large" } */
 /* { dg-final { scan-assembler "globl\t__divti3" } } */
 
 __int128 a, b; void foo () { a = a / b; }
-- 
2.30.2

Re: [r14-5930 Regression] FAIL: gcc.c-torture/compile/libcall-2.c -Os (test for excess errors) on Linux/x86_64

[Jakub Jelinek]

On Wed, Nov 29, 2023 at 07:51:15AM +0100, Jose E. Marchesi wrote:
> > FAIL: gcc.c-torture/compile/libcall-2.c   -O1  (test for excess errors)
> > FAIL: gcc.c-torture/compile/libcall-2.c -O2 -flto
> > -fno-use-linker-plugin -flto-partition=none (test for excess errors)
> > FAIL: gcc.c-torture/compile/libcall-2.c -O2 -flto -fuse-linker-plugin
> > -fno-fat-lto-objects (test for excess errors)
> > FAIL: gcc.c-torture/compile/libcall-2.c   -O2  (test for excess errors)
> > FAIL: gcc.c-torture/compile/libcall-2.c   -O3 -g  (test for excess errors)
> > FAIL: gcc.c-torture/compile/libcall-2.c   -Os  (test for excess errors)
> 
> Sorry about the regression.
> 
> I installed the patch below to skip the test if the target is not x86_64
> in lp64, as obvious.  This should fix the issue.
> 
> >From 4ed0740c6e807460ce79a351094329fdeb551545 Mon Sep 17 00:00:00 2001
> From: "Jose E. Marchesi" <jose.marchesi@oracle.com>
> Date: Wed, 29 Nov 2023 07:44:59 +0100
> Subject: [PATCH] testsuite: fix gcc.c-torture/compile/libcall-2.c in -m32
> 
> This test relies on having __int128 in x86_64 targets, which is only
> available in -m64.
> 
> gcc/testsuite/ChangeLog
> 
> 	* gcc.c-torture/compile/libcall-2.c: Skip test in -m32.
> ---
>  gcc/testsuite/gcc.c-torture/compile/libcall-2.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/gcc/testsuite/gcc.c-torture/compile/libcall-2.c b/gcc/testsuite/gcc.c-torture/compile/libcall-2.c
> index b33944c83ff..9b889172025 100644
> --- a/gcc/testsuite/gcc.c-torture/compile/libcall-2.c
> +++ b/gcc/testsuite/gcc.c-torture/compile/libcall-2.c
> @@ -2,6 +2,8 @@
>     indirect calls.  */
>  
>  /* { dg-do compile } */
> +/* __int128 is not supported in x86 -m32.  */
> +/* { dg-skip-if "" { ! { x86_64-*-* && { ! ilp32 } } } } */
>  /* { dg-options "-O2 -mcmodel=large" { target x86_64-*-* } } */
>  /* { dg-final { scan-assembler "globl\t__divti3" } } */

This is not correct.
When a test uses __int128, it should be guarded with int128 effective
target.
But, as the test doesn't really test anything on non-x86 nor ia32,
it doesn't belong to gcc.c-torture/compile/ at all, it is x86 specific
test which should be moved to gcc.target/i386/libcall-1.c
And, should have
/* { dg-do compile { target int128 } } */
/* { dg-options "-O2 -mcmodel=large" } */
/* { dg-final { scan-assembler "globl\t__divti3" } } */
I guess no need to bother with the extra guard for -mcmodel=large,
because -m32/-mx32 don't have __int128 support, and x86_64-*-*
is incorrect anyway (because with that target one can have all
of -m32/-m64/-mx32).

	Jakub

Re: [r14-5930 Regression] FAIL: gcc.c-torture/compile/libcall-2.c -Os (test for excess errors) on Linux/x86_64

[Jose E. Marchesi]

> On Linux/x86_64,
>
> f31a019d1161ec78846473da743aedf49cca8c27 is the first bad commit
> commit f31a019d1161ec78846473da743aedf49cca8c27
> Author: Jose E. Marchesi <jose.marchesi@oracle.com>
> Date:   Fri Nov 24 06:30:28 2023 +0100
>
>     Emit funcall external declarations only if actually used.
>
> caused
>
> FAIL: gcc.c-torture/compile/libcall-2.c   -O0  (test for excess errors)
> FAIL: gcc.c-torture/compile/libcall-2.c   -O1  (test for excess errors)
> FAIL: gcc.c-torture/compile/libcall-2.c -O2 -flto
> -fno-use-linker-plugin -flto-partition=none (test for excess errors)
> FAIL: gcc.c-torture/compile/libcall-2.c -O2 -flto -fuse-linker-plugin
> -fno-fat-lto-objects (test for excess errors)
> FAIL: gcc.c-torture/compile/libcall-2.c   -O2  (test for excess errors)
> FAIL: gcc.c-torture/compile/libcall-2.c   -O3 -g  (test for excess errors)
> FAIL: gcc.c-torture/compile/libcall-2.c   -Os  (test for excess errors)

Sorry about the regression.

I installed the patch below to skip the test if the target is not x86_64
in lp64, as obvious.  This should fix the issue.

From 4ed0740c6e807460ce79a351094329fdeb551545 Mon Sep 17 00:00:00 2001
From: "Jose E. Marchesi" <jose.marchesi@oracle.com>
Date: Wed, 29 Nov 2023 07:44:59 +0100
Subject: [PATCH] testsuite: fix gcc.c-torture/compile/libcall-2.c in -m32

This test relies on having __int128 in x86_64 targets, which is only
available in -m64.

gcc/testsuite/ChangeLog

	* gcc.c-torture/compile/libcall-2.c: Skip test in -m32.
---
 gcc/testsuite/gcc.c-torture/compile/libcall-2.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/gcc/testsuite/gcc.c-torture/compile/libcall-2.c b/gcc/testsuite/gcc.c-torture/compile/libcall-2.c
index b33944c83ff..9b889172025 100644
--- a/gcc/testsuite/gcc.c-torture/compile/libcall-2.c
+++ b/gcc/testsuite/gcc.c-torture/compile/libcall-2.c
@@ -2,6 +2,8 @@
    indirect calls.  */
 
 /* { dg-do compile } */
+/* __int128 is not supported in x86 -m32.  */
+/* { dg-skip-if "" { ! { x86_64-*-* && { ! ilp32 } } } } */
 /* { dg-options "-O2 -mcmodel=large" { target x86_64-*-* } } */
 /* { dg-final { scan-assembler "globl\t__divti3" } } */
 
-- 
2.30.2

[r14-5930 Regression] FAIL: gcc.c-torture/compile/libcall-2.c -Os (test for excess errors) on Linux/x86_64

[haochen.jiang]

On Linux/x86_64,

f31a019d1161ec78846473da743aedf49cca8c27 is the first bad commit
commit f31a019d1161ec78846473da743aedf49cca8c27
Author: Jose E. Marchesi <jose.marchesi@oracle.com>
Date:   Fri Nov 24 06:30:28 2023 +0100

    Emit funcall external declarations only if actually used.

caused

FAIL: gcc.c-torture/compile/libcall-2.c   -O0  (test for excess errors)
FAIL: gcc.c-torture/compile/libcall-2.c   -O1  (test for excess errors)
FAIL: gcc.c-torture/compile/libcall-2.c   -O2 -flto -fno-use-linker-plugin -flto-partition=none  (test for excess errors)
FAIL: gcc.c-torture/compile/libcall-2.c   -O2 -flto -fuse-linker-plugin -fno-fat-lto-objects  (test for excess errors)
FAIL: gcc.c-torture/compile/libcall-2.c   -O2  (test for excess errors)
FAIL: gcc.c-torture/compile/libcall-2.c   -O3 -g  (test for excess errors)
FAIL: gcc.c-torture/compile/libcall-2.c   -Os  (test for excess errors)

with GCC configured with

../../gcc/configure --prefix=/export/users/haochenj/src/gcc-bisect/master/master/r14-5930/usr --enable-clocale=gnu --with-system-zlib --with-demangler-in-ld --with-fpmath=sse --enable-languages=c,c++,fortran --enable-cet --without-isl --enable-libmpx x86_64-linux --disable-bootstrap

To reproduce:

$ cd {build_dir}/gcc && make check RUNTESTFLAGS="compile.exp=gcc.c-torture/compile/libcall-2.c --target_board='unix{-m32}'"
$ cd {build_dir}/gcc && make check RUNTESTFLAGS="compile.exp=gcc.c-torture/compile/libcall-2.c --target_board='unix{-m32\ -march=cascadelake}'"

(Please do not reply to this email, for question about this report, contact me at haochen dot jiang at intel.com.)
(If you met problems with cascadelake related, disabling AVX512F in command line might save that.)
(However, please make sure that there is no potential problems with AVX512.)