Created attachment 207100 [details] [review]
gobject: use #pragmas to avoid deprecated function warnings

Review of attachment 207100 [details] [review]:

for g_value_array_insert(), we can use an internal function instead of the pragma.

Created attachment 207101 [details] [review]
value-array: Use an internal insert variant

To avoid the deprecation warning.

Review of attachment 207101 [details] [review]:

Looks good - though I'm not sure if it really needs to be inline

I want to keep testing deprecated API - duplicating a copy of it inside the test cases both results in duplicate code *and* not testing API that we still ship.

Review of attachment 207101 [details] [review]:

sorry, I usually inline internal functions - habit from Clutter.

(In reply to comment #5)
> I want to keep testing deprecated API - duplicating a copy of it inside the
> test cases both results in duplicate code *and* not testing API that we still
> ship.

the internal function is inside GValueArray, not the test case; it avoids GValueArray calling a deprecated symbol internally. the public API uses the internal function - as it was always the case.

Review of attachment 207100 [details] [review]:

We could maybe use G_DEFINE_DESTRUCTOR instead of g_atexit for the gobject case ?

Created attachment 207244 [details] [review]
gobject: Use a destructor rather than g_atexit() for refcount debugging

Created attachment 207254 [details] [review]
Add G_GNUC_BEGIN/END_IGNORE_DEPRECATIONS

Add new macros to disable -Wdeprecated-declarations around a piece of
code, using the C99 (and GNU89) _Pragma() operator. Replace the
existing use of #pragma for this in gio, and suppress the warnings in
gvaluearray.c as well.

=====

Here's another approach, which is nice because it provides a nice
macro you can use outside of glib too.

_Pragma() is a C99 addition, but gcc appears to support it even with
"-ansi -pedantic". (At least, current gcc does, and older gccs don't
matter because they don't implement "#pragma diagnostic push/pop" so
we don't define the macro there anyway.)

Review of attachment 207254 [details] [review]:

looks definitely nicer to me

Review of attachment 207254 [details] [review]:

::: glib/docs.c
@@ +1998,3 @@
+ *
+ * This macro can be used either inside or outside of a function body,
+ * but must appear on a line by itself.

I actually got complaints about the pragma in side the function, from people using old gccs. So the last sentence may not really be true.

I also think it would be good to add some guidance here - this is just meant to be a temporary porting tool, adding these guards does not 'fix' deprecations...

Review of attachment 207244 [details] [review]:

::: gobject/gobject.c
@@ +348,3 @@
     }
 }
+#endif /* G_HAS_CONSTRUCTORS */

If I read this correctly, the entire debug_objects_atexit function is inside the ifdef here ? Then what is the g_atexit call in the !G_HAS_CONSTRUCTORS case going to use ?

(In reply to comment #12)
> I actually got complaints about the pragma in side the function, from people
> using old gccs.

Yeah, I saw that, but it turns out that those earlier versions don't implement the diagnostic pragmas anyway, so now we just make the macro only expand to the _Pragma call on recent-enough gccs.

(In reply to comment #13)
> If I read this correctly, the entire debug_objects_atexit function is inside
> the ifdef here ? Then what is the g_atexit call in the !G_HAS_CONSTRUCTORS case
> going to use ?

Oops, yeah, originally I made it so debug objects was only supported now if G_HAS_CONSTRUCTORS, but then I realized that I could just keep using g_atexit() in the non-constructor case, but forgot to fix the ifdef.

Created attachment 207267 [details] [review]
gobject: Use a destructor rather than g_atexit() for refcount debugging

fix !G_HAS_CONSTRUCTORS case

Created attachment 207268 [details] [review]
Add G_GNUC_BEGIN/END_IGNORE_DEPRECATIONS

Added to docs:
 * This is
 * useful for when you have one deprecated function calling another
 * one, or when you still have regression tests for deprecated
 * functions.

Review of attachment 207267 [details] [review]:

looks fine

Review of attachment 207268 [details] [review]:

sure

Attachment 207267 [details] pushed as ab59739 - gobject: Use a destructor rather than g_atexit() for refcount debugging
Attachment 207268 [details] pushed as ca05902 - Add G_GNUC_BEGIN/END_IGNORE_DEPRECATIONS