[Bug libfortran/82962] New: valgrind reports "Conditional jump or move depends on uninitialised value" in EXECUTE_COMMAND_LINE

[janus at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82962

            Bug ID: 82962
           Summary: valgrind reports "Conditional jump or move depends on
                    uninitialised value" in EXECUTE_COMMAND_LINE
           Product: gcc
           Version: 8.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: libfortran
          Assignee: unassigned at gcc dot gnu.org
          Reporter: janus at gcc dot gnu.org
  Target Milestone: ---

Test case:

integer :: istat
! istat = 0
call execute_command_line('echo "Hello World!"', exitstat=istat)
end


After compiling this with gfortran, and running the resulting executable
through valgrind, I get:

==16313== Conditional jump or move depends on uninitialised value(s)
==16313==    at 0x4F3B7EA: _gfortran_execute_command_line_i4 (in
/usr/lib/x86_64-linux-gnu/libgfortran.so.3.0.0)



This only occurs if an EXITSTAT argument is passed, which is not initialized
before the call. However, EXITSTAT is supposed to be an output argument AFAIK.
I don't see why its value should be used at all inside of EXECUTE_COMMAND_LINE?

A dump of the program shows:

MAIN__ ()
{
  integer(kind=4) istat;

  _gfortran_execute_command_line_i4 (&"echo \"Hello World!\""[1]{lb: 1 sz: 1},
0B, &istat, 0B, 0B, 19, 0);
}


Looking at libgfortran/intrinsics/execute_command_line.c, I see the following
in execute_command_line_i4:


  if (exitstat)
    estat_initial = estat = *exitstat;

  execute_command_line (command, w, &estat, cmdstat ? &cstat : NULL,
                        cmdmsg, command_len, cmdmsg_len);

  if (exitstat && estat != estat_initial)
    *exitstat = estat;


Apparenly the existat after the call is compared to the initial exitstat, which
can be uninitialized. I guess it would be better to set exitstat to some
defined initial value, instead of using the input value from outside.