Re: Master is broken

[Andy Moreton]

On Sat 10 Oct 2015, Eli Zaretskii wrote:

>> From: Andy Moreton <address@hidden>
>> Date: Sat, 10 Oct 2015 11:44:24 +0100
>> 
>> > I'm not really sure what's going on here, all this INLINE stuff is too
>> > complicated.  These 2 inline functions are clearly visible in the
>> > preprocessed source, and still the linker barfs.  The only way I could
>> > make it link successfully was by including puresize.h in emacs.c as
>> > well, but then temacs crashed during loadup.
>> 
>> The comments in the definition of INLINE explain how this is supposed to
>> work.
>
> They do?  Then I'm probably missing something, because I cannot find
> where it says that emacs.c should include any header that uses INLINE.
> (Do you mean the comments in conf_post.h?)  I arrived at that by using
> "nm -A" on the various *.o files that use INLINE, looking for *.o
> files that define those symbols.

Yes, the comments in conf_post.h, which I assume come from a gnulib
module. That shows that one file should contain '#define INLINE
EXTERN_INLINE' to ensure that one translation unit contains a
non-inlined definition of each function declared INLINE.

>> I see the same errors with mingw64: the patch below fixed
>> bootstrap for me.
>
> I installed the change in emacs.c recently.  The change in puresize.h
> is not needed to fix the problem, neither on GNU/Linux nor in the
> mingw32 build.  Are you saying that without the changes in puresize.h
> you cannot compile/link the current master?  If so, please show the
> error messages.

Bootstrap of commit f655d09fd5b4 works for me, so the other changes appear
not to be needed. Paul Eggert will know what the correct fix should be.

    AndyM