Re: emacs-28 build failure - gnulib

Re: emacs-28 build failure - gnulib

[Manuel Uberti]

I just pulled 3ff29babbfa2f8b46848944629c1e047291e916b from master and I did:

- git clean -xfd
- ./autogen.sh
- ./configure --with-harfbuzz --with-native-compilation
- make bootstrap -j16

But I am still getting this:

make[2]: Leaving directory '/home/manuel/emacs/doc/lispref'
   CCLD     temacs
/usr/bin/ld: comp.o: in function `comp_hash_source_file':
/home/manuel/emacs/src/comp.c:725: undefined reference to `md5_stream'
collect2: error: ld returned 1 exit status
make[2]: *** [Makefile:647: temacs] Error 1
make[2]: Leaving directory '/home/manuel/emacs/src'
make[1]: *** [Makefile:449: src] Error 2
make[1]: Leaving directory '/home/manuel/emacs'
make: *** [Makefile:1147: bootstrap] Error 2

-- 
Manuel Uberti
www.manueluberti.eu

Re: emacs-28 build failure - gnulib

[Andrea Corallo via Emacs development discussions.]

Manuel Uberti <manuel.uberti@inventati.org> writes:

> I just pulled 3ff29babbfa2f8b46848944629c1e047291e916b from master and I did:
>
> - git clean -xfd
> - ./autogen.sh
> - ./configure --with-harfbuzz --with-native-compilation
> - make bootstrap -j16
>
> But I am still getting this:
>
> make[2]: Leaving directory '/home/manuel/emacs/doc/lispref'
>   CCLD     temacs
> /usr/bin/ld: comp.o: in function `comp_hash_source_file':
> /home/manuel/emacs/src/comp.c:725: undefined reference to `md5_stream'
> collect2: error: ld returned 1 exit status
> make[2]: *** [Makefile:647: temacs] Error 1
> make[2]: Leaving directory '/home/manuel/emacs/src'
> make[1]: *** [Makefile:449: src] Error 2
> make[1]: Leaving directory '/home/manuel/emacs'
> make: *** [Makefile:1147: bootstrap] Error 2

Hi Manuel,

I think this should have been addressed (see bug#50985).

BR

  Andrea

Re: emacs-28 build failure - gnulib

[Manuel Uberti]

On 06/10/21 10:20, Andrea Corallo wrote:
> Hi Manuel,
> 
> I think this should have been addressed (see bug#50985).

I see. Could it be that the fix has not been ported on master yet?

-- 
Manuel Uberti
www.manueluberti.eu

Re: emacs-28 build failure - gnulib

[Robert Pluim]

>>>>> On Wed, 6 Oct 2021 10:23:25 +0200, Manuel Uberti <manuel.uberti@inventati.org> said:

    Manuel> On 06/10/21 10:20, Andrea Corallo wrote:
    >> Hi Manuel,
    >> I think this should have been addressed (see bug#50985).

    Manuel> I see. Could it be that the fix has not been ported on master yet?

emacs-28 has not yet been merged to master.

I can give it a go if people think itʼs urgent, otherwise you can
temporarily cherry-pick 18308b739a and 4d76765b6c

Robert
-- 

Re: emacs-28 build failure - gnulib

[Manuel Uberti]

On 06/10/21 10:56, Robert Pluim wrote:
> emacs-28 has not yet been merged to master.
> 
> I can give it a go if people think itʼs urgent, otherwise you can
> temporarily cherry-pick 18308b739a and 4d76765b6c

It is not urgent, not in my case at least. Thank you for the tip about 
cherry-picking.

-- 
Manuel Uberti
www.manueluberti.eu

emacs-28 build failure - gnulib

[Rajeev N via Emacs development discussions.]

  CCLD     temacs
/usr/bin/ld: comp.o: in function `comp_hash_source_file':
/home/rajeev/tmp/build/em/emacs-28/src/comp.c:725: undefined reference to `md5_stream'
collect2: error: ld returned 1 exit status

Re: emacs-28 build failure - gnulib

[Lars Ingebrigtsen]

Yes, it seems like --native-compilation builds are failing after the
gnulib update; Paul added to the CCs.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: emacs-28 build failure - gnulib

[Andrea Corallo via Emacs development discussions.]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> Yes, it seems like --native-compilation builds are failing after the
> gnulib update; Paul added to the CCs.

On my machine also --without-native-compilation is ATM not functional.
I'm on 1f4ced47a1 and configured with --without-x
--without-native-compilation.

Below the error I see.

Best Regards

  Andrea

 make -C lib all
 make[1]: Entering directory '/home/akrl/emacs2/lib'
   CC       fchmodat.o
 In file included from fchmodat.c:40:
 ./stdlib.h:783:1: error: wrong number of arguments specified for ‘malloc’ attribute
   783 | _GL_FUNCDECL_SYS (aligned_alloc, void *,
       | ^~~~~~~~~~~~~~~~
 ./stdlib.h:783:1: note: expected between 0 and 0, found 2
 ./stdlib.h:839:1: error: wrong number of arguments specified for ‘malloc’ attribute
   839 | _GL_FUNCDECL_SYS (calloc, void *,
       | ^~~~~~~~~~~~~~~~
 ./stdlib.h:839:1: note: expected between 0 and 0, found 2
 ./stdlib.h:863:1: error: wrong number of arguments specified for ‘malloc’ attribute
   863 | _GL_FUNCDECL_SYS (canonicalize_file_name, char *,
       | ^~~~~~~~~~~~~~~~
 ./stdlib.h:863:1: note: expected between 0 and 0, found 2
 ./stdlib.h:1054:1: error: wrong number of arguments specified for ‘malloc’ attribute
  1054 | _GL_FUNCDECL_SYS (malloc, void *,
       | ^~~~~~~~~~~~~~~~
 ./stdlib.h:1054:1: note: expected between 0 and 0, found 2
 ./stdlib.h:1679:1: error: wrong number of arguments specified for ‘malloc’ attribute
  1679 | _GL_FUNCDECL_SYS (realloc, void *, (void *ptr, size_t size)
       | ^~~~~~~~~~~~~~~~
 ./stdlib.h:1679:1: note: expected between 0 and 0, found 2
 In file included from fchmodat.c:41:
 ./string.h:965:1: error: wrong number of arguments specified for ‘malloc’ attribute
   965 | _GL_FUNCDECL_SYS (strdup, char *,
       | ^~~~~~~~~~~~~~~~
 ./string.h:965:1: note: expected between 0 and 0, found 2
 make[1]: *** [Makefile:84: fchmodat.o] Error 1
 make[1]: Leaving directory '/home/akrl/emacs2/lib'
 make: *** [Makefile:436: lib] Error 2

Re: emacs-28 build failure - gnulib

[Paul Eggert]

On 10/5/21 1:05 AM, Andrea Corallo wrote:
> Lars Ingebrigtsen <larsi@gnus.org> writes:
> 
>> Yes, it seems like --native-compilation builds are failing after the
>> gnulib update; Paul added to the CCs.

I think I've now fixed that on the emacs-28 branch; see:

https://bugs.gnu.org/50985#44


>   make -C lib all
>   make[1]: Entering directory '/home/akrl/emacs2/lib'
>     CC       fchmodat.o
>   In file included from fchmodat.c:40:
>   ./stdlib.h:783:1: error: wrong number of arguments specified for ‘malloc’ attribute
>     783 | _GL_FUNCDECL_SYS (aligned_alloc, void *,
>         | ^~~~~~~~~~~~~~~~
>   ./stdlib.h:783:1: note: expected between 0 and 0, found 2

This looks like detritus from a previous build. I've run into that 
problem myself. Please try 'make bootstrap', or start over from a fresh 
Git checkout.

Re: emacs-28 build failure - gnulib

[Andrea Corallo via Emacs development discussions.]

Paul Eggert <eggert@cs.ucla.edu> writes:

> On 10/5/21 1:05 AM, Andrea Corallo wrote:
>> Lars Ingebrigtsen <larsi@gnus.org> writes:
>> 
>>> Yes, it seems like --native-compilation builds are failing after the
>>> gnulib update; Paul added to the CCs.
>
> I think I've now fixed that on the emacs-28 branch; see:
>
> https://bugs.gnu.org/50985#44
>
>
>>   make -C lib all
>>   make[1]: Entering directory '/home/akrl/emacs2/lib'
>>     CC       fchmodat.o
>>   In file included from fchmodat.c:40:
>>   ./stdlib.h:783:1: error: wrong number of arguments specified for ‘malloc’ attribute
>>     783 | _GL_FUNCDECL_SYS (aligned_alloc, void *,
>>         | ^~~~~~~~~~~~~~~~
>>   ./stdlib.h:783:1: note: expected between 0 and 0, found 2
>
> This looks like detritus from a previous build. I've run into that
> problem myself. Please try 'make bootstrap', or start over from a
> fresh Git checkout.

Hi Paul,

I'm getting this error with:

git clean -xfd && ./autogen.sh && ./configure --without-x --without-native-compilation  && time make bootstrap -j16

ATM I'm on d86b2e59c7, is there something else I can try or look into?

Thanks

  Andrea

Re: emacs-28 build failure - gnulib

[Paul Eggert]

[-- Attachment #1: Type: text/plain, Size: 984 bytes --]

On 10/5/21 9:53 AM, Andrea Corallo wrote:
> I'm getting this error with:
> 
> git clean -xfd && ./autogen.sh && ./configure --without-x --without-native-compilation  && time make bootstrap -j16

Unfortunately 'git clean -xfd' doesn't necessarily clean out all the 
detritus. I just now tried the following on Fedora 34 x86-64 from a 
fresh Git checkout (current emacs-28 commit, namely 
1cd1b2835b5e35562c677c48dcf185bb73af4275):

./autogen.sh && ./configure --without-x --without-native-compilation && 
time make bootstrap -j16

and it worked for me (compressed build log attached). Please try this 
approach. If it works for you, it would suggest that 'make bootstrap' 
has a bug in that it doesn't clean out enough stuff somewhere.

I built on a circa-2010 machine (AMD Phenom II X4 910e) with only 4 
CPUs. It's possible that my machine's lack of parallelism helped, though 
if that's the case it would indicate a missing dependency in a Makefile 
somewhere.

[-- Attachment #2: build-log.txt.gz --]
[-- Type: application/gzip, Size: 28037 bytes --]

Re: emacs-28 build failure - gnulib

[Andreas Schwab]

On Okt 05 2021, Paul Eggert wrote:

> Unfortunately 'git clean -xfd' doesn't necessarily clean out all the
> detritus.

What does it leave behind?

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 7578 EB47 D4E5 4D69 2510  2552 DF73 E780 A9DA AEC1
"And now for something completely different."

Re: emacs-28 build failure - gnulib

[Paul Eggert]

On 10/5/21 11:51 AM, Andreas Schwab wrote:
> What does it leave behind?

Sorry, I don't recall. Perhaps directories?

Re: emacs-28 build failure - gnulib

[Andreas Schwab]

On Okt 05 2021, Paul Eggert wrote:

> On 10/5/21 11:51 AM, Andreas Schwab wrote:
>> What does it leave behind?
>
> Sorry, I don't recall. Perhaps directories?

Nope.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 7578 EB47 D4E5 4D69 2510  2552 DF73 E780 A9DA AEC1
"And now for something completely different."

Re: emacs-28 build failure - gnulib

[Eli Zaretskii]

> From: Lars Ingebrigtsen <larsi@gnus.org>
> Date: Tue, 05 Oct 2021 09:25:43 +0200
> Cc: Rajeev N <rajeev.jnk@sivalik.com>, Paul Eggert <eggert@cs.ucla.edu>
> 
> Yes, it seems like --native-compilation builds are failing after the
> gnulib update; Paul added to the CCs.

Gnulib moved md5_stream to a separate file, so that new file should be
imported from Gnulib.