glibc-2.3.3-51 breaks m4

[Steve G]

Hi,

The latest glibc in rawhide breaks the compile of m4:

stackovf.c:108: warning: `__nothrow__' attribute ignored
stackovf.c: In function `setup_stackovf_trap':
stackovf.c:359: error: parse error before "__attribute__"
make[1]: *** [stackovf.o] Error 1
make: *** [all] Error 1

This seems to be centered on the __P macro which is only used in stackovf.c.

I have filed "glibc breaks the build" bug reports and they languish in
bugzilla.
Does someone involved in glibc know that m4 is now broken due to something in the
headers? (glibc-2.3.3-45 works fine) Is glibc going to get fixed/reverted? Or is
m4 going to be fixed?

Thanks,
-Steve Grubb


		
__________________________________
Do you Yahoo!?
Yahoo! Mail - 50x more storage than other providers!
http://promotions.yahoo.com/new_mail

[Ulrich Drepper]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Steve G wrote:

>  The latest glibc in rawhide breaks the compile of m4: 
m4 is just one of many programs which incorrectly use glibc symbols.
These apps never should have done that and the change I've made exposes
this.  But conceding to the laziness of the developers of those packages
I've implemented the underlying change differently and __P will be back
in its old form soon.

- --
➧ Ulrich Drepper ➧ Red Hat, Inc. ➧ 444 Castro St ➧ Mountain View, CA ❖
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.2.6 (GNU/Linux)
>
> iD8DBQFBRzUb2ijCOnn/RHQRAoohAKCK4UyIIDs2wSdlUDYpIkzxQJJgAwCgnqBd
> 4/E6SfV3C//YFLePKg+z1gM=
> =mCOg
> -----END PGP SIGNATURE-----

[Steve G]

> > The latest glibc in rawhide breaks the compile of m4:
>
> m4 is just one of many programs which incorrectly use glibc symbols.
> These apps never should have done that and the change I've made exposes
> this.   
OK, this is what I was curious about. __P is defined in config.h of m4, but
subsequent headers are pulled in. Some of the glibc headers do an undef of __P
which I think is where the problem lies.

> But conceding to the laziness of the developers of those packages
> I've implemented the underlying change differently and __P will be back
> in its old form soon. 
The patch is trivial. I just wanted to know which way this should be fixed. Do
you want to leave glibc in this form or do you want me to file this in bugzilla
with m4's patch?

Thanks for the quick answer.

-Steve Grubb


		
_______________________________
Do you Yahoo!?
Declare Yourself - Register online to vote today!
http://vote.yahoo.com

[Ulrich Drepper]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Steve G wrote:

>  The patch is trivial. I just wanted to know which way this should be
> fixed. Do
>  you want to leave glibc in this form or do you want me to file this in bugzilla
>  with m4's patch? 
glibc-2.3.3-53 (when released) will have the __P macro in the form m4
and other packages expects it.  glibc itself will simply not use the
macro anymore, at all, so it doesn't matter the optimizations we
introduced through __P are not present.  No need to file any bugs
related to this.

- --
➧ Ulrich Drepper ➧ Red Hat, Inc. ➧ 444 Castro St ➧ Mountain View, CA ❖
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.2.6 (GNU/Linux)
>
> iD8DBQFBR6m22ijCOnn/RHQRAge8AJ47VzRIY++2k9LXDzWuka2IQWdwaQCfWzdj
> 2Z2USg50OdXY9GQeFZ6bAhA=
> =JvUf
> -----END PGP SIGNATURE-----