bug#14812: 24.3; build failure on FreeBSD 9.1

bug#14812: 24.3; build failure on FreeBSD 9.1

[Ulrich Mueller]

Forwarding downstream bug: <https://bugs.gentoo.org/473364>

Building Emacs 24.3 (and also Emacs from BZR as of 2013-06-19)
with gcc 4.6 on FreeBSD 9.1 fails:

x86_64-gentoo-freebsd9.1-gcc -std=gnu99 -c  -Demacs  -I. -I/var/tmp/portage/app-editors/emacs-24.3-r2/work/emacs-24.3/src -I../lib -I/var/tmp/portage/app-editors/emacs-24.3-r2/work/emacs-24.3/src/../lib               -MMD -MF deps/xdisp.d -MP     -O2 -pipe xdisp.c
In file included from lisp.h:25:0,
                 from xdisp.c:278:
../lib/stdalign.h:51:0: warning: "_Alignof" redefined [enabled by default]
/usr/include/sys/cdefs.h:240:0: note: this is the location of the previous definition
../lib/stdalign.h:82:0: warning: "_Alignas" redefined [enabled by default]
/usr/include/sys/cdefs.h:239:0: note: this is the location of the previous definition
xdisp.c: In function ‘message_dolog’:
xdisp.c:9449:21: error: expected specifier-qualifier-list before ‘typedef’
xdisp.c: In function ‘set_message_1’:
xdisp.c:10718:12: error: expected specifier-qualifier-list before ‘typedef’
xdisp.c: In function ‘decode_mode_spec_coding’:
xdisp.c:21379:18: error: expected specifier-qualifier-list before ‘typedef’
gmake[1]: *** [xdisp.o] Error 1
gmake[1]: Leaving directory `/var/tmp/portage/app-editors/emacs-24.3-r2/work/emacs-24.3/src'
gmake: *** [src] Error 2

The full build log can be found at:
<https://473364.bugs.gentoo.org/attachment.cgi?id=350998>

Reportedly, the following patch from FreeBSD ports fixes it:
<http://svnweb.freebsd.org/ports/head/editors/emacs-devel/files/patch-lib_verify.h?revision=309486&view=co&pathrev=309486>

bug#14812: 24.3; build failure on FreeBSD 9.1

[Paul Eggert]

Thanks for the bug report.  From the logs it appears that
there are two problems on FreeBDS 9.1, one with verify.h
and the other with stdalign.h.  I installed fixes that
try to address this problem as Emacs trunk bzr 113322;
can you please give that a try?  Thanks.

bug#14812: 24.3; build failure on FreeBSD 9.1

[Ashish SHUKLA]

[-- Attachment #1.1: Type: text/plain, Size: 427 bytes --]

Hi,

I'm the maintainer of editors/emacs* FreeBSD ports, and I'd couple of patches
which are present in FreeBSD ports tree[1]. For this problem, I'm attaching
the patches.

References:
[1]  http://svnweb.freebsd.org/ports/head/editors/emacs-devel/files/

Thanks
-- 
Ashish SHUKLA

“Maxwell's equations have had a greater impact on human history than any ten
presidents.” ("Carl Sagan")

Sent from my Emacs

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1.2: patch-src_lisp.h --]
[-- Type: text/x-chdr, Size: 255 bytes --]


$FreeBSD$

--- src/lisp.h.orig
+++ src/lisp.h
@@ -21,6 +21,10 @@
 #define EMACS_LISP_H
 
 #include <setjmp.h>
+#ifdef __FreeBSD__
+#undef _Alignof
+#undef _Alignas
+#endif
 #include <stdalign.h>
 #include <stdarg.h>
 #include <stdbool.h>

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1.3: patch-lib_verify.h --]
[-- Type: text/x-chdr, Size: 449 bytes --]


$FreeBSD$

--- lib/verify.h.orig
+++ lib/verify.h
@@ -21,6 +21,11 @@
 # define _GL_VERIFY_H
 
 
+// undefine the _Static_assert definition present in <sys/cdefs.h>
+#ifdef __FreeBSD__
+#undef _Static_assert
+#endif
+
 /* Define _GL_HAVE__STATIC_ASSERT to 1 if _Static_assert works as per C11.
    This is supported by GCC 4.6.0 and later, in C mode, and its use
    here generates easier-to-read diagnostics when verify (R) fails.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

bug#14812: 24.3; build failure on FreeBSD 9.1

[Glenn Morris]

Ashish SHUKLA wrote:

> I'm the maintainer of editors/emacs* FreeBSD ports, and I'd couple of
> patches which are present in FreeBSD ports tree[1]. For this problem,
> I'm attaching the patches.
>
> References:
> [1]  http://svnweb.freebsd.org/ports/head/editors/emacs-devel/files/

Most of those patches look more than 6 months old, which means they
predate Emacs 24.3. Had we been told about them, we could have fixed
Emacs 24.3 before it was released.

bug#14812: 24.3; build failure on FreeBSD 9.1

[Ashish SHUKLA]

[-- Attachment #1: Type: text/plain, Size: 1031 bytes --]

On Tue, 09 Jul 2013 02:40:06 -0400, Glenn Morris <rgm@gnu.org> said:
> Ashish SHUKLA wrote:

>> I'm the maintainer of editors/emacs* FreeBSD ports, and I'd couple of
>> patches which are present in FreeBSD ports tree[1]. For this problem,
>> I'm attaching the patches.
>> 
>> References:
>> [1]  http://svnweb.freebsd.org/ports/head/editors/emacs-devel/files/

> Most of those patches look more than 6 months old, which means they
> predate Emacs 24.3. Had we been told about them, we could have fixed
> Emacs 24.3 before it was released.

Sorry about that, though couple of these patches are actually taken from Emacs
bug reports, except for the two which patches which I posted before, and the
one specific to ARM/SPARC architecture which are specific to FreeBSD, and very
hackish in nature.

-- 
Ashish SHUKLA

“We can factor the number 15 with quantum computers. We can also factor the
number 15 with a dog trained to bark three times.” (Robert Harley, sci.crypt,
2001-12-05)

Sent from my Emacs

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

bug#14812: 24.3; build failure on FreeBSD 9.1

[Wolfgang Jenkner]

On Tue, Jul 09 2013, Ashish SHUKLA wrote:

> On Tue, 09 Jul 2013 02:40:06 -0400, Glenn Morris <rgm@gnu.org> said:
>> Ashish SHUKLA wrote:
>
>>> I'm the maintainer of editors/emacs* FreeBSD ports, and I'd couple of
>>> patches which are present in FreeBSD ports tree[1]. For this problem,
>>> I'm attaching the patches.
>>> 
>>> References:
>>> [1]  http://svnweb.freebsd.org/ports/head/editors/emacs-devel/files/
>
>> Most of those patches look more than 6 months old, which means they
>> predate Emacs 24.3. Had we been told about them, we could have fixed
>> Emacs 24.3 before it was released.

But the root problem is that emacs is compiled with -std=gnu99 and
FreeBSD therefore uses its own versions of _Alignas and friends because
of the following test in sys/cdefs.h:

#elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 201112L
/* Do nothing.  They are language keywords. */
#else
/* Not supported.  Implement them using our versions. */
#define	_Alignas(x)		__aligned(x)

Unfortunately this doesn't work but I don't know if this is because of
the way emacs uses this stuff or because of a FreeBSD bug.

Wolfgang

bug#14812: 24.3; build failure on FreeBSD 9.1

[Paul Eggert]

On 07/09/2013 07:04 AM, Wolfgang Jenkner wrote:
> I don't know if this is because of
> the way emacs uses this stuff or because of a FreeBSD bug.

I suspect that the root problem is that FreeBSD 9.1 lacks
a working stdalign.h.  Perhaps this is because they're
implementing C11 gradually and added _Alignof before
adding stdalign.h.  Or maybe their stdalign.h fails
the 'configure' test for stdalign.

When implementing that part of Emacs I assumed that
implementations that had _Alignof would also implement
stdalign.h, but FreeBSD appears to be an exception.
The change in trunk bzr 113322 removes the assumption
that _Alignof is missing if stdalign.h is missing, so
I hope it fixes this porting glitch.

Strictly speaking this would be a bug in FreeBSD only if they're
saying they support C11.  If they claim only C99 support they're
allowed to define _Alignof without providing stdalign.h.  But
in the long run they should provide stdalign.h so the problem
would go away even without the bzr 113322 fix.

I don't know whether the fix (once it's tested)
is worth backporting to Emacs 24.

bug#14812: 24.3; build failure on FreeBSD 9.1

[Wolfgang Jenkner]

On Tue, Jul 09 2013, Paul Eggert wrote:

> On 07/09/2013 07:04 AM, Wolfgang Jenkner wrote:
>> I don't know if this is because of
>> the way emacs uses this stuff or because of a FreeBSD bug.
>
> I suspect that the root problem is that FreeBSD 9.1 lacks
> a working stdalign.h.  Perhaps this is because they're
> implementing C11 gradually and added _Alignof before
> adding stdalign.h.  Or maybe their stdalign.h fails
> the 'configure' test for stdalign.

Sorry, I didn't know that you had already taken care of the bug, and
thanks for the explanation!  I'm a bit confused, though, because we are
talking about a problem which only (IIRC) happens when compiling emacs
with newer versions of gcc, which provide stdalign.h themselves, and the
configure test actually succeeds.  Unfortunately, I can't test things
right now on FreeBSD.

Wolfgang

bug#14812: 24.3; build failure on FreeBSD 9.1

[Glenn Morris]

Paul Eggert wrote:

> I don't know whether the fix (once it's tested) is worth backporting
> to Emacs 24.

It seems unlikely there will be another release from emacs-24,
so I don't think you need to bother (unless it is totally trivial).

bug#14812: 24.3; build failure on FreeBSD 9.1

[Ulrich Mueller]

>>>>> On Sun, 07 Jul 2013, Paul Eggert wrote:

> Thanks for the bug report.  From the logs it appears that
> there are two problems on FreeBDS 9.1, one with verify.h
> and the other with stdalign.h.  I installed fixes that
> try to address this problem as Emacs trunk bzr 113322;
> can you please give that a try?  Thanks.

Thank you, this appears to fix it.
Tested on FreeBSD 9.1 with GCC 4.6.4.

Ulrich

bug#14812: 24.3; build failure on FreeBSD 9.1

[Paul Eggert]

On 07/13/2013 02:35 AM, Ulrich Mueller wrote:

> Thank you, this appears to fix it.

Thanks, marking it as done.

bug#14812: 24.3; build failure on FreeBSD 9.1

[Glenn Morris]

Ashish SHUKLA wrote:

>>> [1]  http://svnweb.freebsd.org/ports/head/editors/emacs-devel/files/
[...]
> Sorry about that, though couple of these patches are actually taken
> from Emacs bug reports, except for the two which patches which I
> posted before, and the one specific to ARM/SPARC architecture which
> are specific to FreeBSD, and very hackish in nature.

If there are any patches there that you think we need in Emacs trunk,
and don't yet have, could you point them out? (Ideally, with a separate
bug report for each.) The patches don't seem to have clear descriptions
of what they do.

The configure.ac patches removes the test for ALSA altogether - why?
And adds some stuff related to processes (?), see below.

You don't need any of the patch-doc-*-Makefile.in any more,
since --no-split is the default since 24.3.

I guess
patch-src_coding.c == http://debbugs.gnu.org/8522 ?
If you've tested it and it works, it would be helpful to add that
comment in that report.

The src/lisp.h, lib/verify.h ones are redundant now that this report is
fixed?

The sources.el one is some packaging-specific thing?

The src/sysdep.c one is to do with list-system-processes, but Emacs 24
already has an implementation for FreeBSD, so I don't understand.
The src/process.c and part of the configure.ac patch are related to this?

bug#14812: 24.3; build failure on FreeBSD 9.1

[Ashish SHUKLA]

Hi Glenn,

Thanks for going through the diffs.

On Tue, 16 Jul 2013 14:58:24 -0400, Glenn Morris <rgm@gnu.org> said:
> Ashish SHUKLA wrote:

>>>> [1]  http://svnweb.freebsd.org/ports/head/editors/emacs-devel/files/
> [...]
>> Sorry about that, though couple of these patches are actually taken
>> from Emacs bug reports, except for the two which patches which I
>> posted before, and the one specific to ARM/SPARC architecture which
>> are specific to FreeBSD, and very hackish in nature.

> If there are any patches there that you think we need in Emacs trunk,
> and don't yet have, could you point them out? (Ideally, with a separate
> bug report for each.) The patches don't seem to have clear descriptions
> of what they do.

> The configure.ac patches removes the test for ALSA altogether - why?

ALSA is not native sound system for FreeBSD, and OSS support is already
available so it was excluded.

> And adds some stuff related to processes (?), see below.

the libutil.h change (in src/process.c) was for openpty which Emacs was using
at one point of time[1]. It'll be removed as it's not needed anymore.

> You don't need any of the patch-doc-*-Makefile.in any more,
> since --no-split is the default since 24.3.

I'll remove it as well.

> I guess
> patch-src_coding.c == http://debbugs.gnu.org/8522 ?
> If you've tested it and it works, it would be helpful to add that
> comment in that report.

I'll ask original reporter to reply to that thread, as I'm not a Japanese
user.

> The src/lisp.h, lib/verify.h ones are redundant now that this report is
> fixed?

Yes, I'll remove them, once I make sure they don't break anything with all
compilers FreeBSD is targeting (clang, GCC 4.2, GCC 4.6+).

> The sources.el one is some packaging-specific thing?

They're for installing sources with Emacs, borrowed from Gentoo GNU/Linux patch.

> The src/sysdep.c one is to do with list-system-processes, but Emacs 24
> already has an implementation for FreeBSD, so I don't understand.
> The src/process.c and part of the configure.ac patch are related to this?

This patch was from bug# 5243, which seems to be committed last year, so it's
not needed anymore.

Are the changes for ARM/SPARC architecture from patch-src_sysdep.c good enough ?
If you think those are fine, then I can file a bug report along with the changes.

References:
[1]  http://www.freebsd.org/cgi/query-pr.cgi?pr=161181

Thanks
-- 
Ashish SHUKLA

”A verbal contract isn't worth the paper it's written on.” (Samuel Goldwyn)

Sent from my Emacs

bug#14812: 24.3; build failure on FreeBSD 9.1

[Glenn Morris]

Ashish SHUKLA wrote:

>> The configure.ac patches removes the test for ALSA altogether - why?
>
> ALSA is not native sound system for FreeBSD, and OSS support is already
> available so it was excluded.

I don't quite understand. ALSA may be present on a FreeBSD system, and
if it is, configure detecting it actually breaks Emacs sound support, so
it should be excluded?

If the test for ossaudio succeeds, we should not go on to test for alsa
too? Sounds like this might be general rather than specific to FreeBSD?

> Are the changes for ARM/SPARC architecture from patch-src_sysdep.c
> good enough ?
> If you think those are fine, then I can file a bug report along with
> the changes.

I don't know. It's certainly worth making an Emacs bug report that
explains what the problem is and how the patch is supposed to fix it.

bug#14812: 24.3; build failure on FreeBSD 9.1

[Ashish SHUKLA]

[-- Attachment #1: Type: text/plain, Size: 1349 bytes --]

On Fri, 19 Jul 2013 20:57:37 -0400, Glenn Morris <rgm@gnu.org> said:
> Ashish SHUKLA wrote:

>>> The configure.ac patches removes the test for ALSA altogether - why?
>> 
>> ALSA is not native sound system for FreeBSD, and OSS support is already
>> available so it was excluded.

> I don't quite understand. ALSA may be present on a FreeBSD system, and
> if it is, configure detecting it actually breaks Emacs sound support, so
> it should be excluded?

> If the test for ossaudio succeeds, we should not go on to test for alsa
> too? Sounds like this might be general rather than specific to FreeBSD?

In configure, ALSA test is done before falling back to OSS. That's why I've
excluded as OSS is native on FreeBSD.

>> Are the changes for ARM/SPARC architecture from patch-src_sysdep.c
>> good enough ?
>> If you think those are fine, then I can file a bug report along with
>> the changes.

> I don't know. It's certainly worth making an Emacs bug report that
> explains what the problem is and how the patch is supposed to fix it.

Great. I'll file a bug report later today for these changes.

Thanks
-- 
Ashish SHUKLA

“There was truth and there was untruth, and if you clung to the truth even
against the whole world, you were not mad.” (George Orwell, "Nineteen
Eighty-Four", 1949)

Sent from my Emacs

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

bug#14812: 24.3; build failure on FreeBSD 9.1

[Jan Djärv]

Hello.

20 jul 2013 kl. 07:21 skrev ashish.is@lostca.se (Ashish SHUKLA):

> In configure, ALSA test is done before falling back to OSS. That's why I've
> excluded as OSS is native on FreeBSD.

You should never remove things like this if you want patches to be applied.  Just make something like:

if ! freebsd
  test for alsa
end if

Test for ALSA must preceed OSS on GNU/Linux and possibly other OS:es.

	Jan D.

bug#14812: 24.3; build failure on FreeBSD 9.1

[Andreas Schwab]

ashish.is@lostca.se (Ashish SHUKLA) writes:

> In configure, ALSA test is done before falling back to OSS. That's why I've
> excluded as OSS is native on FreeBSD.

What would go wrong if one would want to use alsa on freebsd?

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#14812: 24.3; build failure on FreeBSD 9.1

[Ashish SHUKLA]

[-- Attachment #1: Type: text/plain, Size: 754 bytes --]

On Sat, 20 Jul 2013 11:20:40 +0200, Andreas Schwab <schwab@linux-m68k.org> said:
> ashish.is@lostca.se (Ashish SHUKLA) writes:

>> In configure, ALSA test is done before falling back to OSS. That's why I've
>> excluded as OSS is native on FreeBSD.

> What would go wrong if one would want to use alsa on freebsd?

Well, I tried sometime ago, and it didn't work for me. And there is no way in
Emacs, to specify OSS over ALSA, esp. when OSS works fine and native. I
decided to disable ALSA support. I never felt the need of it, nor anyone asked
for it.

HTH
-- 
Ashish SHUKLA

“How can I possibly put a new idea into your heads, if I do not first remove
your delusions?” (Robert A. Heinlein, "Life-Line", 1939)

Sent from my Emacs

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

bug#14812: 24.3; build failure on FreeBSD 9.1

[Andreas Schwab]

ashish.is@lostca.se (Ashish SHUKLA) writes:

> Well, I tried sometime ago, and it didn't work for me. And there is no way in
> Emacs, to specify OSS over ALSA, esp. when OSS works fine and native. I
> decided to disable ALSA support. I never felt the need of it, nor anyone asked
> for it.

I'd expect that those who install alsa on their freebsd system know what
they are doing.  Otherwise the configure check should do no harm.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#14812: 24.3; build failure on FreeBSD 9.1

[Glenn Morris]

Ashish SHUKLA wrote:

> And there is no way in Emacs, to specify OSS over ALSA, esp. when OSS
> works fine and native. I decided to disable ALSA support.

We could extend configure's --with-sound option to allow
--with-sound=oss or --with-sound=alsa, if that is a useful thing to do,
but "no-one's ever asked for that".

> I never felt the need of it, nor anyone asked for it.

bug#14812: 24.3; build failure on FreeBSD 9.1

[Ashish SHUKLA]

[-- Attachment #1: Type: text/plain, Size: 737 bytes --]

On Sat, 20 Jul 2013 15:43:29 -0400, Glenn Morris <rgm@gnu.org> said:
> Ashish SHUKLA wrote:

>> And there is no way in Emacs, to specify OSS over ALSA, esp. when OSS
>> works fine and native. I decided to disable ALSA support.

> We could extend configure's --with-sound option to allow
> --with-sound=oss or --with-sound=alsa, if that is a useful thing to do,
> but "no-one's ever asked for that".

I prefer this approach, otherwise ALSA will be picked if user has installed
ALSA libraries but has no intention to build ALSA support in Emacs.

Thanks
-- 
Ashish SHUKLA

“I look at life as being cruise director on the Titanic. I may not get there,
but I'm going first class.” (Art Buchwald)

Sent from my Emacs

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

bug#14812: 24.3; build failure on FreeBSD 9.1

[Ashish SHUKLA]

[-- Attachment #1: Type: text/plain, Size: 483 bytes --]

On Sat, 20 Jul 2013 10:51:30 +0530, ashish.is@lostca.se (Ashish SHUKLA) said:

[...]


> Great. I'll file a bug report later today for these changes.

Filed bug report #14923[1] which you must have seen already.

References:
[1]  http://debbugs.gnu.org/cgi/bugreport.cgi?bug=14923

Thanks
-- 
Ashish SHUKLA

“The only difference between the saint and the sinner is that every saint has a
past and every sinner has a future.” (Oscar Wilde)

Sent from my Emacs

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

bug#14812: 24.3; build failure on FreeBSD 9.1

[Andrew W. Nosenko]

On Sat, Jul 20, 2013 at 2:31 PM, Andreas Schwab <schwab@linux-m68k.org> wrote:
> ashish.is@lostca.se (Ashish SHUKLA) writes:
>
>> Well, I tried sometime ago, and it didn't work for me. And there is no way in
>> Emacs, to specify OSS over ALSA, esp. when OSS works fine and native. I
>> decided to disable ALSA support. I never felt the need of it, nor anyone asked
>> for it.
>
> I'd expect that those who install alsa on their freebsd system know what
> they are doing.  Otherwise the configure check should do no harm.

Unfortunatelly, ALSA may be installed just as consequence/dependency
of some other port or package.  IIRC, some time ago Firefox had broken
OSS support (it just crashed) forcing to have ALSA installed as
unwanted consequence, just for example.

-- 
Andrew W. Nosenko <andrew.w.nosenko@gmail.com>

bug#14812: 24.3; build failure on FreeBSD 9.1

[Glenn Morris]

>> We could extend configure's --with-sound option to allow
>> --with-sound=oss or --with-sound=alsa, if that is a useful thing to do,

Implemented. (Some of configure's logic in this area doesn't make sense
to me.)