[Bug tree-optimization/98444] New: [10 Regression] compile error with -ftracer and -Werror=format-overflow

[Bug tree-optimization/98444] New: [10 Regression] compile error with -ftracer and -Werror=format-overflow

[drfiemost at email dot it]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98444

            Bug ID: 98444
           Summary: [10 Regression] compile error with -ftracer and
                    -Werror=format-overflow
           Product: gcc
           Version: 10.2.1
               URL: https://github.com/elogind/elogind/issues/189
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: drfiemost at email dot it
  Target Milestone: ---

Created attachment 49843
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=49843&action=edit
Preprocessed source

I get the following error compiling the attached code with gcc-10 (gcc version
10.2.1 20201205):

$ gcc -O2 -ftracer -Werror=format-overflow -c pam_elogind.i 
../src/login/pam_elogind.c: In function ‘export_legacy_dbus_address’:
../src/login/pam_elogind.c:315:13: error: ‘%s’ directive argument is null
[-Werror=format-overflow=]
cc1: some warnings being treated as errors


According to upstream [1] the variable is asserted before calling the function
so it will never be null.
The code compiles fine with gcc-9 and even with gcc-10 when removing the
-ftracer flag.
I can't tell if it's a regression in gcc-10, just a false positive or what so
I'm reporting it here.

[1] https://github.com/elogind/elogind/issues/189

[Bug tree-optimization/98444] [10 Regression] compile error with -ftracer and -Werror=format-overflow

[msebor at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98444

Martin Sebor <msebor at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |msebor at gcc dot gnu.org
   Last reconfirmed|                            |2020-12-27
             Status|UNCONFIRMED                 |NEW
      Known to work|                            |11.0, 9.3.0
     Ever confirmed|0                           |1
      Known to fail|                            |10.2.0

--- Comment #1 from Martin Sebor <msebor at gcc dot gnu.org> ---
The warning is based on the IL below where the %s argument is null, so it's
working correctly.  It doesn't consider the complex control flow (bb2 -> b10 ->
bb5) from which it could deduce the asprintf call with the null pointer is only
reachable conditionally to issue a more nuanced message, but that wouldn't
prevent it, only make its conditional nature more apparent (all warnings are
conditional on the function they're in being called).  The assertion that the
runtime pointer is nonnull is in export_legacy_dbus_address's caller,
configure_runtime_directory, 
which is inlined into its caller, and has no effect on the code in
export_legacy_dbus_address.  Adding something like:

  if (!runtime) __builtin_unreachable ();

just before the problematic call to asprintf() avoids the warning.

In GCC 9 or on trunk (GCC 11), jump threading doesn't introduce the the invalid
call so the warning doesn't trigger.  So I can confirm this regression for GCC
10 but I don't expect to be able to do anything about it there.  Longer term,
we're aware of these warnings for synthesized code but we're still looking for
a solution to avoid them.

export_legacy_dbus_address (struct pam_handle_t * handle, const char * runtime)
{
  ...
  <bb 2> [local count: 397250656]:
  t = 0B;
  if (runtime_22(D) != 0B)
    goto <bb 3>; [94.50%]
  else
    goto <bb 10>; [5.50%]

  <bb 3> [local count: 382684072]:
  _65 = strlen (runtime_22(D));
  _n__16 = _65 + 5;
  if (_n__16 > 4194304)
    goto <bb 4>; [10.58%]
  else
    goto <bb 11>; [89.42%]

  <bb 4> [local count: 39725066]:
  log_assert_failed_realm (0, "sizeof(char)*_n_ <= ALLOCA_MAX",
&"../src/login/pam_elogind.c"[3], 311, &__PRETTY_FUNCTION__);

  <bb 5> [local count: 7591956]:
  _10 = asprintf (&t, "unix:path=%s/bus", 0B);   <<< warning here
  if (_10 < 0)
    goto <bb 6>; [26.36%]
  else
    goto <bb 7>; [73.64%]

  ...
  <bb 10> [local count: 21848788]:
  _29 = __builtin_alloca (1);
  *_29 = 0;
  _appendees_ ={v} {CLOBBER};
  _9 = access (_29, 0);
  if (_9 < 0)
    goto <bb 9>; [42.09%]
  else
    goto <bb 5>; [57.91%]

[Bug tree-optimization/98444] [10 Regression] compile error with -ftracer and -Werror=format-overflow

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98444

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |10.3

[Bug tree-optimization/98444] [10 Regression] compile error with -ftracer and -Werror=format-overflow

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98444

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|10.3                        |10.4

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 10.3 is being released, retargeting bugs to GCC 10.4.

[Bug tree-optimization/98444] [10 Regression] compile error with -ftracer and -Werror=format-overflow

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98444

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|10.4                        |10.5

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
GCC 10.4 is being released, retargeting bugs to GCC 10.5.

[Bug tree-optimization/98444] [10 Regression] compile error with -ftracer and -Werror=format-overflow

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98444

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to fail|                            |10.5.0
             Status|NEW                         |RESOLVED
   Target Milestone|10.5                        |11.0
         Resolution|---                         |FIXED

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed in GCC 11.