[Bug jit/110466] New: jit.dg FAILs on ppc64le

[Bug jit/110466] New: jit.dg FAILs on ppc64le

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110466

            Bug ID: 110466
           Summary: jit.dg FAILs on ppc64le
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: jit
          Assignee: dmalcolm at gcc dot gnu.org
          Reporter: mpolacek at gcc dot gnu.org
  Target Milestone: ---

On ppc64le:

$ make check RUNTESTFLAGS=jit.exp

Running /home/polacek/src/gcc/gcc/testsuite/jit.dg/jit.exp ... 
FAIL: jit.dg/test-combination.c, initial compilation
FAIL: FAIL
FAIL: FAIL
FAIL: FAIL
FAIL: jit.dg/test-expressions.c, initial compilation
FAIL: jit.dg/test-threads.c, initial compilation

                === jit Summary === 

# of expected passes            6632
# of unexpected failures        6
# of unsupported tests          4
make[1]: Leaving directory `/home/polacek/x/trunk/gcc'

I suspect that's so since

commit d2e782cb99c3116c389d6a9565678c4ffe267777
Author: Antoni Boucher <bouanto@zoho.com>
Date:   Sun Nov 20 10:22:53 2022 -0500

    libgccjit: Fix float vector comparison

[Bug jit/110466] jit.dg FAILs on ppc64le

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110466

--- Comment #1 from Marek Polacek <mpolacek at gcc dot gnu.org> ---
The problem is that the patch above uses __vector which is wrong on Power. 
However, with my patch:

--- a/gcc/testsuite/jit.dg/test-expressions.c
+++ b/gcc/testsuite/jit.dg/test-expressions.c
@@ -560,17 +560,17 @@ verify_comparisons (gcc_jit_result *result)
   CHECK_VALUE (test_COMPARISON_GE_on_int (1, 2), 0);
   CHECK_VALUE (test_COMPARISON_GE_on_int (2, 1), 1);

-  typedef int __vector __attribute__ ((__vector_size__ (sizeof(int) * 2)));
-  typedef __vector (*test_vec_fn) (__vector, __vector);
+  typedef int v2si __attribute__ ((__vector_size__ (sizeof(int) * 2)));
+  typedef v2si (*test_vec_fn) (v2si, v2si);

-  __vector zero_zero = {0, 0};
-  __vector zero_one = {0, 1};
-  __vector one_zero = {1, 0};
+  v2si zero_zero = {0, 0};
+  v2si zero_one = {0, 1};
+  v2si one_zero = {1, 0};

-  __vector true_true = {-1, -1};
-  __vector false_true = {0, -1};
-  __vector true_false = {-1, 0};
-  __vector false_false = {0, 0};
+  v2si true_true = {-1, -1};
+  v2si false_true = {0, -1};
+  v2si true_false = {-1, 0};
+  v2si false_false = {0, 0};

   test_vec_fn test_COMPARISON_EQ_on_vec_int =
     (test_vec_fn)gcc_jit_result_get_code (result,
@@ -615,7 +615,7 @@ verify_comparisons (gcc_jit_result *result)
   CHECK_VECTOR_VALUE (2, test_COMPARISON_GE_on_vec_int (zero_one, one_zero),
false_true);

   typedef float __vector_f __attribute__ ((__vector_size__ (sizeof(float) *
2)));
-  typedef __vector (*test_vec_f_fn) (__vector_f, __vector_f);
+  typedef v2si (*test_vec_f_fn) (__vector_f, __vector_f);

   __vector_f zero_zero_f = {0, 0};
   __vector_f zero_one_f = {0, 1};

I see FAILs like

Running /home/polacek/src/gcc/gcc/testsuite/jit.dg/jit.exp ... 
FAIL: test-combination.c.exe iteration 1 of 5: verify_comparisons: actual:
test_COMPARISON_EQ_on_vec_int (zero_zero, zero_zero) != expected: true_true
(position 0)
FAIL: test-combination.c.exe killed: 7835 exp8 0 0 CHILDKILLED SIGABRT SIGABRT
FAIL: FAIL
FAIL: FAIL
FAIL: FAIL
FAIL: test-expressions.c.exe iteration 1 of 5: verify_comparisons: actual:
test_COMPARISON_EQ_on_vec_int (zero_zero, zero_zero) != expected: true_true
(position 0)
FAIL: test-expressions.c.exe killed: 10718 exp12 0 0 CHILDKILLED SIGABRT
SIGABRT
FAIL: test-threads.c.exe: verify_comparisons: actual:
test_COMPARISON_EQ_on_vec_int (zero_zero, zero_zero) != expected: true_true
(position 0)
FAIL: test-threads.c.exe killed: 12273 exp12 0 0 CHILDKILLED SIGABRT SIGABRT

and I don't know what that means.

[Bug jit/110466] jit.dg FAILs on ppc64le

[dmalcolm at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110466

David Malcolm <dmalcolm at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2023-06-28
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |ASSIGNED

--- Comment #2 from David Malcolm <dmalcolm at gcc dot gnu.org> ---
Confirmed (on cfarm 112)

I was able to get test-expressions.c to fully pass on that box with this
further change:

diff --git a/gcc/testsuite/jit.dg/test-expressions.c
b/gcc/testsuite/jit.dg/test-expressions.c
index de6022f..2337b01 100644
--- a/gcc/testsuite/jit.dg/test-expressions.c
+++ b/gcc/testsuite/jit.dg/test-expressions.c
@@ -417,7 +417,7 @@ static void run_test_of_comparison(gcc_jit_context *ctxt,
                         const char *expected)
 {
   gcc_jit_type *vec_type =
-    gcc_jit_type_get_vector (type, 4);
+    gcc_jit_type_get_vector (type, 2);

   CHECK_STRING_VALUE (
     make_test_of_comparison (ctxt,

[Bug jit/110466] jit.dg FAILs on ppc64le

[dmalcolm at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110466

--- Comment #3 from David Malcolm <dmalcolm at gcc dot gnu.org> ---
With the above fix, the remaining failures are:

FAIL: FAIL
FAIL: FAIL
FAIL: FAIL

which are from testdebuginfo.c; I see:

  Reading symbols from
/home/dmalcolm/build/gcc/testsuite/jit4/jit-debuginfo.o...Dwarf Error: wrong
version in compilation unit header (is 5, should be 2, 3, or 4) [in module
/home/dmalcolm/build/gcc/testsuite/jit4/jit-debuginfo.o]

though I'm not sure if that's the cause of the error.

[Bug jit/110466] jit.dg FAILs on ppc64le

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110466

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to David Malcolm from comment #3)
> 
>   Reading symbols from
> /home/dmalcolm/build/gcc/testsuite/jit4/jit-debuginfo.o...Dwarf Error: wrong
> version in compilation unit header (is 5, should be 2, 3, or 4) [in module
> /home/dmalcolm/build/gcc/testsuite/jit4/jit-debuginfo.o]
> 
> though I'm not sure if that's the cause of the error.

Those look depedenent on the binutils version that is in use. I suspect if
there is a newer version of binutils installed, it would just work. Dwarf5 was
added to binutils version 2.35.2 . Though maybe it is finding the wrong readelf
...

[Bug jit/110466] jit.dg FAILs on ppc64le

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110466

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #4)
> (In reply to David Malcolm from comment #3)
> > 
> >   Reading symbols from
> > /home/dmalcolm/build/gcc/testsuite/jit4/jit-debuginfo.o...Dwarf Error: wrong
> > version in compilation unit header (is 5, should be 2, 3, or 4) [in module
> > /home/dmalcolm/build/gcc/testsuite/jit4/jit-debuginfo.o]
> > 
> > though I'm not sure if that's the cause of the error.
> 
> Those look depedenent on the binutils version that is in use. I suspect if
> there is a newer version of binutils installed, it would just work. Dwarf5
> was added to binutils version 2.35.2 . Though maybe it is finding the wrong
> readelf ...

Or rather it is gdb that is complaining and a newer gdb version is needed.

[Bug jit/110466] jit.dg FAILs on ppc64le

[dmalcolm at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110466

--- Comment #6 from David Malcolm <dmalcolm at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #5)
> (In reply to Andrew Pinski from comment #4)
> > (In reply to David Malcolm from comment #3)
> > > 
> > >   Reading symbols from
> > > /home/dmalcolm/build/gcc/testsuite/jit4/jit-debuginfo.o...Dwarf Error: wrong
> > > version in compilation unit header (is 5, should be 2, 3, or 4) [in module
> > > /home/dmalcolm/build/gcc/testsuite/jit4/jit-debuginfo.o]
> > > 
> > > though I'm not sure if that's the cause of the error.
> > 
> > Those look depedenent on the binutils version that is in use. I suspect if
> > there is a newer version of binutils installed, it would just work. Dwarf5
> > was added to binutils version 2.35.2 . Though maybe it is finding the wrong
> > readelf ...
> 
> Or rather it is gdb that is complaining and a newer gdb version is needed.

Yeah; I think the gdb version is too old.

Adding the following patch turns this into an "unsupported" (on this host, at
least):

diff --git a/gcc/testsuite/jit.dg/jit.exp b/gcc/testsuite/jit.dg/jit.exp
index 3568dbb..8bf7e51 100644
--- a/gcc/testsuite/jit.dg/jit.exp
+++ b/gcc/testsuite/jit.dg/jit.exp
@@ -440,6 +440,10 @@ proc jit-check-debug-info { obj_file cmds match } {
         send $cmd
     }
     expect {
+       -re "Dwarf Error: wrong version in compilation unit header" {
+           set testcase [testname-for-summary]
+           unsupported "$testcase: gdb does not support dwarf version"
+       }
         -re $match { pass OK }
         default { fail FAIL }
     }

I'll try the combination of these patches on my regular x86_64 workstation...

[Bug jit/110466] jit.dg FAILs on ppc64le

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110466

--- Comment #7 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by David Malcolm <dmalcolm@gcc.gnu.org>:

https://gcc.gnu.org/g:6735d66083953315c0d7d491df07d36061093a24

commit r14-2222-g6735d66083953315c0d7d491df07d36061093a24
Author: David Malcolm <dmalcolm@redhat.com>
Date:   Fri Jun 30 11:20:02 2023 -0400

    jit: avoid using __vector in testcase [PR110466]

    r13-4531-gd2e782cb99c311 added test coverage to libgccjit's vector
    support, but used __vector, which doesn't work on Power.  Additionally
    the size param to gcc_jit_type_get_vector was wrong.

    Fixed thusly.

    gcc/testsuite/ChangeLog:
            PR jit/110466
            * jit.dg/test-expressions.c (run_test_of_comparison): Fix size
            param to gcc_jit_type_get_vector.
            (verify_comparisons): Use a typedef rather than __vector.

    Co-authored-by: Marek Polacek <polacek@redhat.com>
    Signed-off-by: David Malcolm <dmalcolm@redhat.com>

[Bug jit/110466] jit.dg FAILs on ppc64le

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110466

--- Comment #8 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by David Malcolm <dmalcolm@gcc.gnu.org>:

https://gcc.gnu.org/g:c3c0ba5436170e01499f4390b7b628a32943a9e2

commit r14-2223-gc3c0ba5436170e01499f4390b7b628a32943a9e2
Author: David Malcolm <dmalcolm@redhat.com>
Date:   Fri Jun 30 11:20:02 2023 -0400

    jit.exp: handle dwarf version mismatch in jit-check-debug-info [PR110466]

    gcc/testsuite/ChangeLog:
            PR jit/110466
            * jit.dg/jit.exp (jit-check-debug-info): Gracefully handle too
            early versions of gdb that don't support our dwarf version, via
            "unsupported".

    Signed-off-by: David Malcolm <dmalcolm@redhat.com>

[Bug jit/110466] jit.dg FAILs on ppc64le

[dmalcolm at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110466

--- Comment #9 from David Malcolm <dmalcolm at gcc dot gnu.org> ---
Should be fixed on trunk for gcc 14 by the above commits.

Keeping open to track backporting to gcc 13.

[Bug jit/110466] jit.dg FAILs on ppc64le

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110466

--- Comment #10 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-13 branch has been updated by David Malcolm
<dmalcolm@gcc.gnu.org>:

https://gcc.gnu.org/g:b38472ffc1e631bd357573b44d956ce16d94e666

commit r13-8744-gb38472ffc1e631bd357573b44d956ce16d94e666
Author: David Malcolm <dmalcolm@redhat.com>
Date:   Thu May 9 13:09:27 2024 -0400

    jit: avoid using __vector in testcase [PR110466]

    r13-4531-gd2e782cb99c311 added test coverage to libgccjit's vector
    support, but used __vector, which doesn't work on Power.  Additionally
    the size param to gcc_jit_type_get_vector was wrong.

    Fixed thusly.

    gcc/testsuite/ChangeLog:
            PR jit/110466
            * jit.dg/test-expressions.c (run_test_of_comparison): Fix size
            param to gcc_jit_type_get_vector.
            (verify_comparisons): Use a typedef rather than __vector.

    (cherry picked from commit r14-2222-g6735d660839533)

    Co-authored-by: Marek Polacek <polacek@redhat.com>
    Signed-off-by: David Malcolm <dmalcolm@redhat.com>

[Bug jit/110466] jit.dg FAILs on ppc64le

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110466

--- Comment #11 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-13 branch has been updated by David Malcolm
<dmalcolm@gcc.gnu.org>:

https://gcc.gnu.org/g:60dcb710b6b4aa22ea96abc8df6dfe9067f3d7fe

commit r13-8745-g60dcb710b6b4aa22ea96abc8df6dfe9067f3d7fe
Author: David Malcolm <dmalcolm@redhat.com>
Date:   Thu May 9 13:09:27 2024 -0400

    jit.exp: handle dwarf version mismatch in jit-check-debug-info [PR110466]

    gcc/testsuite/ChangeLog:
            PR jit/110466
            * jit.dg/jit.exp (jit-check-debug-info): Gracefully handle too
            early versions of gdb that don't support our dwarf version, via
            "unsupported".

    (cherry picked from commit r14-2223-gc3c0ba5436170e)

    Signed-off-by: David Malcolm <dmalcolm@redhat.com>

[Bug jit/110466] jit.dg FAILs on ppc64le

[dmalcolm at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110466

David Malcolm <dmalcolm at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
             Status|ASSIGNED                    |RESOLVED

--- Comment #12 from David Malcolm <dmalcolm at gcc dot gnu.org> ---
Should be fixed for GCC 13 (for the upcoming GCC 13.3) by the above patch.