[Bug c++/94185] New: gcc-10 crashes with "error: unable to generate reloads for {*zero_extendsidi2} internal compiler error: in curr_insn_transform, at lra-constraints.c:4006

[Bug c++/94185] New: gcc-10 crashes with "error: unable to generate reloads for {*zero_extendsidi2} internal compiler error: in curr_insn_transform, at lra-constraints.c:4006

[slyfox at inbox dot ru]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94185

            Bug ID: 94185
           Summary: gcc-10 crashes with "error: unable to generate reloads
                    for {*zero_extendsidi2} internal compiler error: in
                    curr_insn_transform, at lra-constraints.c:4006
           Product: gcc
           Version: 10.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: slyfox at inbox dot ru
  Target Milestone: ---

Created attachment 48037
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=48037&action=edit
bug.cc

Originally observed on smartmontools-7.0, reduced with creduce down to:

struct a {
  int b;
  int c();
  a() : b(c()) {}
  ~a();
  char *e();
};
struct f {
  void g(int);
};
struct ar {
  int au[256];
  f h(int);
} bb;
a i();
a j(int);
long k(int, ar);
int d;
void l(char *, ar m, long n) {
  switch (m.au[d])
  case 0:
    n &= 4294967295;
  bb.h(0).g(n);
}
void o() {
  ar bd;
  a bh, bi, attrname = j(0) = i();
  int be = k(0, bd);
  l(attrname.e(), bd, be);
}

Exact command to crash gcc-master:

$ ./xg++ -B. -O2 -c -o /tmp/bug.o /tmp/bug.cc -fstack-protector-strong -fPIE
/tmp/bug.cc: In function 'void o()':
/tmp/bug.cc:30:1: error: unable to generate reloads for:
   30 | }
      | ^
(insn 62 61 63 8 (set (mem/c:DI (plus:DI (reg/f:DI 19 frame)
                (const_int -2152 [0xfffffffffffff798])) [7 %sfp+-2104 S8 A64])
        (zero_extend:DI (mem/c:SI (plus:DI (reg/f:DI 19 frame)
                    (const_int -2152 [0xfffffffffffff798])) [7 %sfp+-2104 S4
A64]))) "/tmp/bug.cc":22:7 114 {*zero_extendsidi2}
     (nil))
during RTL pass: reload
/tmp/bug.cc:30:1: internal compiler error: in curr_insn_transform, at
lra-constraints.c:4006
0x7fc532a94dd9 __libc_start_main
        ../csu/libc-start.c:308
Please submit a full bug report,
with preprocessed source if appropriate.
Please include the complete backtrace with any bug report.
See <https://gcc.gnu.org/bugs/> for instructions.

Compiler is built as:

$ ./xg++ -B. -v
Reading specs from ./specs
COLLECT_GCC=./xg++
COLLECT_LTO_WRAPPER=./lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: ../gcc/configure --enable-languages=c,c++ --disable-bootstrap
--with-multilib-list=m64
--prefix=/home/slyfox/dev/git/gcc-native-quick/../gcc-native-quick-installed
--disable-nls --without-isl --disable-libsanitizer --disable-libvtv
--disable-libgomp --disable-libstdcxx-pch --disable-libunwind-exceptions
CFLAGS='-O1 ' CXXFLAGS='-O1 ' --with-sysroot=/usr/x86_64-HEAD-linux-gnu
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 10.0.1 20200315 (experimental) (GCC)

[Bug target/94185] [10 Regression] crashes with "error: unable to generate reloads for {*zero_extendsidi2} internal compiler error: in curr_insn_transform, at lra-constraints.c:4006

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94185

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2020-03-16
   Target Milestone|---                         |10.0
            Summary|gcc-10 crashes with "error: |[10 Regression] crashes
                   |unable to generate reloads  |with "error: unable to
                   |for {*zero_extendsidi2}     |generate reloads for
                   |internal compiler error: in |{*zero_extendsidi2}
                   |curr_insn_transform, at     |internal compiler error: in
                   |lra-constraints.c:4006      |curr_insn_transform, at
                   |                            |lra-constraints.c:4006
           Priority|P3                          |P1
           Keywords|                            |ra
      Known to work|                            |9.3.0
     Ever confirmed|0                           |1

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Must be a recent regression.

[Bug target/94185] [10 Regression] crashes with "error: unable to generate reloads for {*zero_extendsidi2} internal compiler error: in curr_insn_transform, at lra-constraints.c:4006

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94185

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org,
                   |                            |vmakarov at gcc dot gnu.org

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Indeed, started with r10-7165-ga4504f32c056db781a2bdc104dffa1b29684c930

[Bug target/94185] [10 Regression] crashes with "error: unable to generate reloads for {*zero_extendsidi2} internal compiler error: in curr_insn_transform, at lra-constraints.c:4006

[zsojka at seznam dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94185

--- Comment #3 from Zdenek Sojka <zsojka at seznam dot cz> ---
I am hitting probably the same issue:
$ cat testcase.c
int a, b, c;

int
foo (char e, short f, long g)
{
  g = __builtin_mul_overflow (a, c, &b) ? g : (unsigned) g;
  g &= __builtin_clrsbll (b);
  return e + f + g + a + b + c;
}
$ x86_64-pc-linux-gnu-gcc -O -fno-omit-frame-pointer -mbranch-cost=1 testcase.c
testcase.c: In function 'foo':
testcase.c:9:1: error: unable to generate reloads for:
    9 | }
      | ^
(insn 27 26 28 3 (set (mem/c:DI (plus:DI (reg/f:DI 19 frame)
                (const_int -56 [0xffffffffffffffc8])) [2 %sfp+-8 S8 A64])
        (zero_extend:DI (mem/c:SI (plus:DI (reg/f:DI 19 frame)
                    (const_int -56 [0xffffffffffffffc8])) [2 %sfp+-8 S4 A64])))
114 {*zero_extendsidi2}
     (nil))
during RTL pass: reload
testcase.c:9:1: internal compiler error: in curr_insn_transform, at
lra-constraints.c:4006
0x6d3384 _fatal_insn(char const*, rtx_def const*, char const*, int, char
const*)
        /repo/gcc-trunk/gcc/rtl-error.c:108
0x68fcfe curr_insn_transform
        /repo/gcc-trunk/gcc/lra-constraints.c:4006
0xe36e76 lra_constraints(bool)
        /repo/gcc-trunk/gcc/lra-constraints.c:5025
0xe2199c lra(_IO_FILE*)
        /repo/gcc-trunk/gcc/lra.c:2437
0xdd5479 do_reload
        /repo/gcc-trunk/gcc/ira.c:5523
0xdd5479 execute
        /repo/gcc-trunk/gcc/ira.c:5709
Please submit a full bug report,
with preprocessed source if appropriate.
Please include the complete backtrace with any bug report.
See <https://gcc.gnu.org/bugs/> for instructions.

and also on another testcase; -fno-omit-frame-pointer seems to be the common
compiler switch

[Bug target/94185] [10 Regression] crashes with "error: unable to generate reloads for {*zero_extendsidi2} internal compiler error: in curr_insn_transform, at lra-constraints.c:4006

[vmakarov at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94185

--- Comment #4 from Vladimir Makarov <vmakarov at gcc dot gnu.org> ---
Thank you the reduced test.  I've started to work on this.

[Bug target/94185] [10 Regression] crashes with "error: unable to generate reloads for {*zero_extendsidi2} internal compiler error: in curr_insn_transform, at lra-constraints.c:4006

[vmakarov at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94185

--- Comment #5 from Vladimir Makarov <vmakarov at gcc dot gnu.org> ---
I found the problem.  LRA reused the same insn alternative when mem subreg was
changed.

The patch will be ready today or tomorrow at worst.

[Bug target/94185] [10 Regression] crashes with "error: unable to generate reloads for {*zero_extendsidi2} internal compiler error: in curr_insn_transform, at lra-constraints.c:4006

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94185

--- Comment #6 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Vladimir Makarov <vmakarov@gcc.gnu.org>:

https://gcc.gnu.org/g:bae7b38cf8a21e068ad5c0bab089dedb78af3346

commit r10-7197-gbae7b38cf8a21e068ad5c0bab089dedb78af3346
Author: Vladimir N. Makarov <vmakarov@redhat.com>
Date:   Mon Mar 16 16:42:19 2020 -0400

    Fix PR94185: Do not reuse insn alternative after changing memory subreg.

    2020-03-16  Vladimir Makarov  <vmakarov@redhat.com>

            PR target/94185
            * lra-spills.c (remove_pseudos): Do not reuse insn alternative
            after changing memory subreg.

    2020-03-16  Vladimir Makarov  <vmakarov@redhat.com>

            PR target/94185
            * g++.target/i386/pr94185.C: New test.

[Bug target/94185] [10 Regression] crashes with "error: unable to generate reloads for {*zero_extendsidi2} internal compiler error: in curr_insn_transform, at lra-constraints.c:4006

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94185

Martin Liška <marxin at gcc dot gnu.org> changed:

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

--- Comment #7 from Martin Liška <marxin at gcc dot gnu.org> ---
I hope we can close it.

[Bug target/94185] [10 Regression] crashes with "error: unable to generate reloads for {*zero_extendsidi2} internal compiler error: in curr_insn_transform, at lra-constraints.c:4006

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94185

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

https://gcc.gnu.org/g:994d48620621fa33d32018be5fb70042e38546d5

commit r10-7209-g994d48620621fa33d32018be5fb70042e38546d5
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Tue Mar 17 11:12:59 2020 +0100

    testsuite: Fix pr94185.C testcase on i686-linux with C++98 [PR94185]

    I'm getting on i686-linux
    FAIL: g++.target/i386/pr94185.C  -std=gnu++98 (test for excess errors)
    This is because of a diagnostic that 4294967295 is unsigned only in ISO
C90.
    Adding U suffix fixes it and the testcase still ICEs with unfixed gcc and
    passes with current trunk.

    2020-03-17  Jakub Jelinek  <jakub@redhat.com>

            PR target/94185
            * g++.target/i386/pr94185.C (l): Use 4294967295U instead of
4294967295
            to avoid FAIL with -m32 -std=c++98.

[Bug target/94185] [10 Regression] crashes with "error: unable to generate reloads for {*zero_extendsidi2} internal compiler error: in curr_insn_transform, at lra-constraints.c:4006

[slyfox at inbox dot ru]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94185

--- Comment #9 from Sergei Trofimovich <slyfox at inbox dot ru> ---
Now smartmontools-7.0 builds successfully against gcc version 10.0.1 20200317
(experimental), commit 2e30d3e3e88b6a544074ff89de4974bc5e200e89.

Thank you for a superfast fix!