[Xcb] [PATCH] v2: Link xcb-util* with -no-undefined

[Jon TURNEY]

On 25/03/2011 17:11, Gaetan Nadon wrote:
> On Fri, 2011-03-25 at 12:41 +0000, Jon TURNEY wrote:
>> I'm not sure what other common linker flags might ever be used, but as
>> requested:
> 
> If you grep the xorg libraries, there are no other flag. So this is
> over-kill.
> 
>>
>> A series of patches for the split xcb-util repositories adding a
>> variable
>> XCB_UTIL_COMMON_LDFLAGS to xcb_util_common.m4, and using that in all
>> xcb-util
>> library Makefiles.  Set XCB_UTIL_COMMON_LDFLAGS to '-no-undefined'.
> 
> These patches only compile on your workstation. They do not contain the
> change in
> the m4 submodule to point to the new commit in common-m4.
> The current commit number can be seen here:
> http://cgit.freedesktop.org/xcb/util/tree/?repo=m4&page=commit&id=177e4afb9b57a7b7589c15d8a9bd85b4a5524128
> Don't worry about the "invalid commit reference".

Ah, I see. So that's how git submodules work :-)

I'm not going to propose patches which contain guesses as to what the id of a
commit will be when it appears in the master repository, though.

If an xcb committer wants apply the xcb-util-common-m4 patch, I'll produce
updated patches for the other repositories. (I managed to make a mistake in
the xcb-util-renderutil patch in any case)

In the alternative, some guidance as what's need to an acceptable patch which
achieves the goal of adding -no-undefined to the linker flags would be welcome.

> Each of the containing repo  must have "m4" to be updated to pint at the
> commit provided by the common-m4
> patch. It's best to test it in  a separate directory starting clean.