[Bug c++/106862] New: ice in compute_control_dep_chain, at gimple-predicate-analysis.cc:1105

[Bug c++/106862] New: ice in compute_control_dep_chain, at gimple-predicate-analysis.cc:1105

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106862

            Bug ID: 106862
           Summary: ice in compute_control_dep_chain, at
                    gimple-predicate-analysis.cc:1105
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: dcb314 at hotmail dot com
  Target Milestone: ---

Created attachment 53545
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=53545&action=edit
gzipped C++ source code

For the attached C++ code, recent g++ does this:

$ /home/dcb/gcc/results/bin/g++ -c -g -O1 -Wall -fpermissive bug842.cc
during GIMPLE pass: uninit
In file included from VSLRead.C:113:
y.tab.c: In function ‘int VSLLib_parse()’:
y.tab.c:1497:1: internal compiler error: in compute_control_dep_chain, at
gimple-predicate-analysis.cc:1105
0x1df8e69 compute_control_dep_chain(basic_block_def*, basic_block_def const*,
vec<edge_def*, va_heap, vl_ptr>*, uns
igned int*, vec<edge_def*, va_heap, vl_ptr>&, unsigned int*, unsigned int,
unsigned int)
        ../../trunk.git/gcc/gimple-predicate-analysis.cc:1105
0x1df8ae7 compute_control_dep_chain(basic_block_def*, basic_block_def const*,
vec<edge_def*, va_heap, vl_ptr>*, uns
igned int*, vec<edge_def*, va_heap, vl_ptr>&, unsigned int*, unsigned int,
unsigned int)
        ../../trunk.git/gcc/gimple-predicate-analysis.cc:1086
0x1df8ae7 compute_control_dep_chain(basic_block_def*, basic_block_def const*,
vec<edge_def*, va_heap, vl_ptr>*, uns
igned int*, vec<edge_def*, va_heap, vl_ptr>&, unsigned int*, unsigned int,
unsigned int)
        ../../trunk.git/gcc/gimple-predicate-analysis.cc:1086
0x1df7895 compute_control_dep_chain(basic_block_def*, basic_block_def const*,
vec<edge_def*, va_heap, vl_ptr>*, uns
igned int*, unsigned int)
        ../../trunk.git/gcc/gimple-predicate-analysis.cc:1120

The bug first seems to occur sometime between git hash 5e070cf4bd085e10,
from yesterday, and d6582c662ca0da05, from today.

[Bug tree-optimization/106862] ice in compute_control_dep_chain, at gimple-predicate-analysis.cc:1105

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106862

--- Comment #1 from David Binderman <dcb314 at hotmail dot com> ---
I have a cvise reduction running. 

The code is in C++, which cvise isn't very efficient at reducing, so I expect 
it will be some hours before I have any finished reduction.

[Bug tree-optimization/106862] ice in compute_control_dep_chain, at gimple-predicate-analysis.cc:1105

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106862

--- Comment #2 from David Binderman <dcb314 at hotmail dot com> ---
Reduced C++ code seems to be:

extern fprintf(char);
struct string {
  string(char);
};
struct {
  call();
  replace(string);
} vsllib;
struct ListNode {
  ListNode();
};
struct TestNode {
  TestNode(int);
};
typedef struct {
  *node;
} YYSTYPE;
int vsldebug, VSLLib_parse_vsln, VSLLib_parse_vslvsp_4294967293_0_0;
VSLLib_parse() {
  YYSTYPE *vslvsp;
vslnewstate:
  switch (VSLLib_parse_vsln) {
  case 13:
    vslvsp = 0;
  case 6:
    string("");
  case 35:
    delete vslvsp;
    VSLLib_parse_vslvsp_4294967293_0_0 &&vslvsp[0].node ? new ListNode : 0;
  case 72:
    delete vslvsp;
  case 75:
    vslvsp ? vsllib.call() : 0;
  case 9:
    vsllib.call();
  case 1:
    vsllib.call();
  case 3:
    vsllib.call();
  case 5:
    vsllib.call();
  case 7:
    vsllib.call();
  case 0:
    vsllib.call();
  case 2:
    VSLLib_parse_vslvsp_4294967293_0_0 &&vslvsp[0].node ? new TestNode(new int)
                                                        : 0;
    string func_name = vsllib.replace(func_name) && vsllib.replace(0);
    delete vslvsp;
  }
  ++vslvsp;
  if (vsldebug)
    fprintf("");
  goto vslnewstate;
}

[Bug tree-optimization/106862] ice in compute_control_dep_chain, at gimple-predicate-analysis.cc:1105

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106862

--- Comment #3 from David Binderman <dcb314 at hotmail dot com> ---
Another test case, this time in C is:

enum { GELB, VIOLETT } KABEL;
typedef playground_t[][4][2];
int ToPlayground_0_0_0, PlayGame_countdown, PlayGame_count_tick_len,
    PlayGame_move_tick_len, ProcessCapsules_row;
playground_t CapsuleCountdown;
void PlayGame() {
  int FinishTakeover = 0, prev_move_tick;
  while (!FinishTakeover) {
    if (KeyIsPressedR() && ModIsPressed() && ModIsPressed())
      return;
    if (PlayGame_count_tick_len) {
      if (PlayGame_countdown)
        EndCountdownSound();
      FinishTakeover = 1;
    }
    if (prev_move_tick + PlayGame_move_tick_len)
      prev_move_tick += ProcessCapsules();
  }
}
void ProcessCapsules() {
  int color = GELB;
  for (; color <= VIOLETT; color++) {
    ProcessCapsules_row = 0;
    for (; ProcessCapsules_row < 12; ProcessCapsules_row++) {
      if (CapsuleCountdown[color][0][ProcessCapsules_row])
        CapsuleCountdown[color][0][ProcessCapsules_row]--;
      if (CapsuleCountdown[color][0][ProcessCapsules_row])
        ToPlayground_0_0_0 = KABEL;
    }
  }
}

$ /home/dcb/gcc/results/bin/gcc -c -fcommon -O3 -Wall bug842B.c

[Bug tree-optimization/106862] ice in compute_control_dep_chain, at gimple-predicate-analysis.cc:1105

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106862

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |DUPLICATE

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
r13-2493-g1a4e1425f84985 (PR106844) seems to fix this.

*** This bug has been marked as a duplicate of bug 106844 ***