798954 - NS_ALWAYS_INLINE triggers "warning: always_inline function might not be inlinable [-Wattributes]"

Status: RESOLVED DUPLICATE of bug 697810

(Core :: General, defect)

Description

[Daniel Holbert [:dholbert]]

Warning: -Wattributes in /mozilla/toolkit/components/places/SQLFunctions.cpp: always_inline function might not be inlinable
/mozilla/toolkit/components/places/SQLFunctions.cpp:95:3: warning: always_inline function might not be inlinable [-Wattributes]

Warning: -Wattributes in /mozilla/toolkit/components/places/SQLFunctions.cpp: always_inline function might not be inlinable
/mozilla/toolkit/components/places/SQLFunctions.cpp:48:3: warning: always_inline function might not be inlinable [-Wattributes]

Pretty sure these functions need to be declared with "inline" **and** NS_ALWAYS_INLINE if we actually want them to be inlined.  The GCC "always_inline" attribute only has an effect if it's combined with "inline".

From the GCC docs:
> GCC does not inline any functions when not optimizing unless you
> specify the `always_inline' attribute for the function, like this:
>     /* Prototype.  */
>     inline void foo (const char) __attribute__((always_inline));
(note: ^^^^^                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ both are specified.
http://gcc.gnu.org/onlinedocs/gcc/Inline.html

See also this ZFS commit, fixing the same warning in their code, with a commit message explaining this as well:
  https://github.com/zfsonlinux/zfs/commit/10be533e3344f523e1b8d6ab4f0658897a95ac02

Comment 1

[Daniel Holbert [:dholbert]]

(In reply to Daniel Holbert [:dholbert] from comment #0)
> Pretty sure these functions need to be declared with "inline" **and**
> NS_ALWAYS_INLINE if we actually want them to be inlined.  The GCC
> "always_inline" attribute only has an effect if it's combined with "inline".

Hmm, though actually there's a message on the GCC help list that disagrees with this:

  If you use the always_inline attribute, then, yes, you do not need to
  also use the inline keyword.

http://gcc.gnu.org/ml/gcc-help/2007-01/msg00051.html

Still, if I add "inline", the build warning goes away. So I think it's a good idea, even if it's not strictly necessary.

Comment 2

[Daniel Holbert [:dholbert]]

(In reply to Daniel Holbert [:dholbert] from comment #1)
> Hmm, though actually there's a message on the GCC help list that disagrees
> with this:
[...]
> http://gcc.gnu.org/ml/gcc-help/2007-01/msg00051.html

...though that's 5 years old, so I think I trust the newer information / documentation more.

I actually posted to m.d.platform about this:
  https://groups.google.com/forum/?fromgroups=#!topic/mozilla.dev.platform/kAAEDEoQcfM

Morphing this bug to be about the general issue, rather than about the specific instance in SQLFunctions.cpp.

Comment 3

[Daniel Holbert [:dholbert]]

jgilbert points me to bug 697810 which is already filed on this macro being broken.  This is effectively a dupe of that bug.