[Bug target/113076] New: [14] RISC-V: gfortran.dg/dec_io_1.f90 runtime error after r14-4971-g0beb1611754

[Bug target/113076] New: [14] RISC-V: gfortran.dg/dec_io_1.f90 runtime error after r14-4971-g0beb1611754

[ewlu at rivosinc dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113076

            Bug ID: 113076
           Summary: [14] RISC-V: gfortran.dg/dec_io_1.f90 runtime error
                    after r14-4971-g0beb1611754
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: ewlu at rivosinc dot com
  Target Milestone: ---

At line 66 of file ../gcc/gcc/testsuite/gfortran.dg/dec_io_1.f90 (unit = 3,
file = 'dec_io_1.txt')
Fortran runtime error: Problem in CLOSE

Error termination. Backtrace:
#0  0x2b2fa9db in ???
#1  0x2b2fb2f9 in ???
#2  0x2b2fb993 in ???
#3  0x2b38a765 in ???
#4  0x10c41 in MAIN__
        at ../gcc/gcc/testsuite/gfortran.dg/dec_io_1.f90:66
#5  0x10701 in main
        at ../gcc/gcc/testsuite/gfortran.dg/dec_io_1.f90:101


for linux rv32gcv-ilp32d targets.

#3  0x2b38a765 appears to be  // hex values may differ from the gdb outputs
<_gfortran_st_close+210>: ef 30 a7 9b     jal     ra,0x2b2fba30
<_gfortran_generate_error>

compile command:
./build-gcc-linux-stage2/gcc/testsuite/gfortran12/../../gfortran
-B./build-gcc-linux-stage2/gcc/testsuite/gfortran12/../../
-B./build-gcc-linux-stage2/riscv64-unknown-linux-gnu/lib32/ilp32d/libgfortran/
../gcc/gcc/testsuite/gfortran.dg/dec_io_1.f90 -march=rv32gcv -mabi=ilp32d
-mcmodel=medlow -fdiagnostics-plain-output -fdiagnostics-plain-output -O2 -fdec
-B./build-gcc-linux-stage2/riscv64-unknown-linux-gnu/lib32/ilp32d/libgfortran/.libs
-L./build-gcc-linux-stage2/riscv64-unknown-linux-gnu/lib32/ilp32d/libgfortran/.libs
-L./build-gcc-linux-stage2/riscv64-unknown-linux-gnu/lib32/ilp32d/libgfortran/.libs
-L./build-gcc-linux-stage2/riscv64-unknown-linux-gnu/lib32/ilp32d/libatomic/.libs
-lm -g -o ./dec_io_1.exe

run command:
QEMU_CPU=rv32,vlen=128,v=true,vext_spec=v1.0,Zve32f=true,Zve64f=true
./bin/qemu-riscv32 ./dec_io_1.exe

Bisected to r14-4971-g0beb1611754

[Bug target/113076] [14] RISC-V: gfortran.dg/dec_io_1.f90 runtime error after r14-4971-g0beb1611754

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113076

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Target|                            |riscv

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
The bisection is highly suspicious since that's a x86 specific change that
doesn't affect riscv.

[Bug target/113076] [14] RISC-V: gfortran.dg/dec_io_1.f90 runtime error after r14-4972-g8aa47713701

[ewlu at rivosinc dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113076

Edwin Lu <ewlu at rivosinc dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[14] RISC-V:                |[14] RISC-V:
                   |gfortran.dg/dec_io_1.f90    |gfortran.dg/dec_io_1.f90
                   |runtime error after         |runtime error after
                   |r14-4971-g0beb1611754       |r14-4972-g8aa47713701

--- Comment #2 from Edwin Lu <ewlu at rivosinc dot com> ---
(In reply to Richard Biener from comment #1)
> The bisection is highly suspicious since that's a x86 specific change that
> doesn't affect riscv.

Sorry about that I linked the wrong commit. I bisected it down to
r14-4972-g8aa47713701

8aa47713701b1f1878b81169852269a299272e87 is the first bad commit
commit 8aa47713701b1f1878b81169852269a299272e87
Author: Vladimir N. Makarov <vmakarov@redhat.com>
Date:   Fri Oct 27 08:28:24 2023 -0400

    [RA]: Add cost calculation for reg equivalence invariants

    My recent patch improving cost calculation for pseudos with equivalence
    resulted in failure of gcc.target/arm/eliminate.c on aarch64.  This patch
    fixes this failure.

    gcc/ChangeLog:

            * ira-costs.cc: (get_equiv_regno, calculate_equiv_gains):
            Process reg equivalence invariants.

 gcc/ira-costs.cc | 4 ++++
 1 file changed, 4 insertions(+)