[Bug c/81568] attribute always_inline honored even after attribute noinline

[dave.pagan at oracle dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81568

--- Comment #5 from Dave Pagan  ---
Thanks for the update, Martin.

[Bug c/81568] attribute always_inline honored even after attribute noinline

[msebor at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81568

Martin Sebor  changed:

   What|Removed |Added

   See Also||https://gcc.gnu.org/bugzill
   ||a/show_bug.cgi?id=81544

--- Comment #4 from Martin Sebor  ---
Fixed in r255469.

[Bug c/81568] attribute always_inline honored even after attribute noinline

[msebor at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81568

Martin Sebor  changed:

   What|Removed |Added

 Status|UNCONFIRMED |RESOLVED
 Resolution|--- |FIXED
   Assignee|unassigned at gcc dot gnu.org  |msebor at gcc dot 
gnu.org
  Known to fail||5.4.0, 6.4.0, 7.2.0

--- Comment #3 from Martin Sebor  ---
Thanks for the ping!  I just committed a fix for pr81544 (and pr81566) that
makes attribute validation more consistent.  With the patch applied, the output
for the test case in comment #0 is as follows.  There's just one warning and no
error because (as the warning says) the conflicting attribute is dropped.

I think with that this bug can be resolved as fixed as well (I forgot about it
and so didn't mention it in the commit).  Let me go ahead and do that, but
please feel free to take the freshly committed patch for a spin and open new
bugs if/when you find them or if you have suggestions for further refinements
or enhancements in this area.

$ cat z.c && gcc -O2 -S -Wall -Wextra -Wpedantic z.c
int __attribute__ ((noinline)) f (int);
int __attribute__ ((always_inline)) f (int);

int f (int i) { return i > 1 ? i * f (i - 1) * f (i - 2) : i > 0 ? i * f (i -
1) : 1; }

int f1 (void)
{
  return f (123);
}
z.c:2:1: warning: ignoring attribute ‘always_inline’ because it conflicts with
attribute ‘noinline’ [-Wattributes]
 int __attribute__ ((always_inline)) f (int);
 ^~~
z.c:1:32: note: previous declaration here
 int __attribute__ ((noinline)) f (int);
^

[Bug c/81568] attribute always_inline honored even after attribute noinline

[dave.pagan at oracle dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81568

Dave Pagan  changed:

   What|Removed |Added

 CC||dave.pagan at oracle dot com

--- Comment #2 from Dave Pagan  ---
Martin, have you (or has anyone else) spent time on this, or have a fix
already? I was thinking about looking into it further but don't want to
duplicate effort.

[Bug c/81568] attribute always_inline honored even after attribute noinline

[msebor at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81568

Martin Sebor  changed:

   What|Removed |Added

   Keywords||diagnostic, wrong-code
   Severity|normal  |minor

--- Comment #1 from Martin Sebor  ---
The cause of the problem is in the validation for these two attributes being
done in two places: one in handle_always_inline_attribute and
handle_noinline_attribute in c-attribs.c, and another in
diagnose_mismatched_attributes.  The former rejects the attribute if it's found
to conflict with the other on the same declaration, but it doesn't detect
conflicts across declarations.  The latter does detect those conflicts but it
doesn't reject them.  As a result, the same problem gets handled differently in
the two situations.