[Bug gcov-profile/83505] [GCOV] a pointer assignment before a dead return statment lead to incorrect coverage in gcov

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83505

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2017-12-20
                 CC|                            |mpolacek at gcc dot gnu.org
           Assignee|unassigned at gcc dot gnu.org      |marxin at gcc dot 
gnu.org
     Ever confirmed|0                           |1

--- Comment #2 from Martin Liška <marxin at gcc dot gnu.org> ---
Confirmed.

There are some interesting observations I've just done:

G++ FE optimizes out 'if (1)' condition, while C keeps it:

C:
{
  int t = 1;
  int * p = &t;

    int t = 1;
    int * p = &t;
  if (1)
    {
      if (0)
        {
          return 0;
        }
      return 1;
    }
}
return 0;

t = {CLOBBER};
t = {CLOBBER};

C++:
{
  int t = 1;
  int * p = &t;

  <<cleanup_point   int t = 1;>>;
    int * p = &t;
  if (0)
    {
      return <retval> = 0;
    }
  return <retval> = 1;
}
return <retval> = 0;

Which leads to different EH tree creation. In C, we have ndests == 2, thus we
copy BB with clobber of t:
t = {CLOBBER};

Then later on, one of these blocks is removed and the one with location set to
'return 0' wins.
That leads me to question whether we should emit location of clobbers to gcno
files. I have to think about
it more.

Marek is the described FE behavior expected?
Anyway, thanks for the report.