[Bug c/71787] New: #pragma GCC diagnostic unexpectedly terminates a statement

["gary at intrepid dot com" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71787

            Bug ID: 71787
           Summary: #pragma GCC diagnostic unexpectedly terminates a
                    statement
           Product: gcc
           Version: 7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: gary at intrepid dot com
  Target Milestone: ---

This bug may be related to PR41517.

Consider the following program:


#define NULL ((void *)0)

typedef enum {GET, SET} op_t;

extern void abort ();

void
get_set (op_t op,
         int *ref,
         int value __attribute__ ((unused)),
         int *fetch)
{
  int orig_value;
  switch (op)
    {
    case GET:
      orig_value = *ref;
      break;
    case SET:
      if (fetch)
        orig_value = *ref;
      *ref = value;
      break;
    }
  if (fetch)
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wmaybe-uninitialized"
    *fetch = orig_value;
#pragma GCC diagnostic pop
}

int
main ()
{
  int v = 0;
  int value;
  get_set (GET, &v,  1, &value);
  get_set (SET, &v, 10, &value);
  get_set (SET, &v, 20, NULL);
}

When compiled and executed, it segfaults because the diagnostic push statement
is being handled as if it were a semi-colon.

If the sequence of statements above is changed to:

#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wmaybe-uninitialized"
  if (fetch)
    *fetch = orig_value;
#pragma GCC diagnostic pop

It executes as expected.

This is either a bug, or it would be helpful if the documentation is updated to
describe this behavior.