bug#38868: Problem compiling Emacs24 on Debian 10

bug#38868: Problem compiling Emacs24 on Debian 10

[Eduardo Ochs]

Hi list,

I updated my Debian from Stretch (a.k.a. Debian 9) to Buster (a.k.a.
Debian 10) two days ago, and I noticed that in Buster the package
"emacs24" is essentially just a fake package that depends on
emacs-gtk... links:

  https://wiki.debian.org/DebianStretch
  https://wiki.debian.org/DebianBuster
  http://packages.debian.org/stretch/emacs24
  http://packages.debian.org/buster/emacs24

I am the author of a package called "eev" that is in ELPA and that
says this in its eev.el:

  ;; Package-Requires: ((emacs "24.1"))

I wanted to be sure that I wouldn't break compatibility with emacs24
if I changed a part of eev that has a defadvice, so I tried to compile
an emacs24 in my machine (for tests!!!) by doing this:


  # See: https://packages.debian.org/source/buster/emacs
  sudo apt-get build-dep -y emacs

  mkdir -p   ~/bigsrc/
  rm -Rf     ~/bigsrc/emacs24/
  cd         ~/bigsrc/
  git clone git://git.sv.gnu.org/emacs ~/bigsrc/emacs24
  cd         ~/bigsrc/emacs24/

  export PAGER=cat
  git branch --list -a
  git for-each-ref

  git checkout emacs-24

  ./autogen.sh   2>&1 | tee oa
  ./configure    2>&1 | tee oc
  # make         2>&1 | tee om
  make bootstrap 2>&1 | tee omb

  # "make bootstrap" fails!
  tail -n 15 omb
  # The output of "tail" is:
  # make[3]: Leaving directory '/home/edrx/bigsrc/emacs24/lisp'
  # if test "no" = "yes"; then \
  #   rm -f bootstrap-emacs; \
  #   ln temacs bootstrap-emacs; \
  # else \
  #   ./temacs --batch --load loadup bootstrap || exit 1; \
  #   test "X" = X ||  -zex emacs; \
  #   mv -f emacs bootstrap-emacs; \
  # fi
  # /bin/bash: line 7: 20593 Segmentation fault      ./temacs --batch
--load loadup bootstrap
  # make[2]: *** [Makefile:816: bootstrap-emacs] Error 1
  # make[2]: Leaving directory '/home/edrx/bigsrc/emacs24/src'
  # make[1]: *** [Makefile:387: src] Error 2
  # make[1]: Leaving directory '/home/edrx/bigsrc/emacs24'
  # make: *** [Makefile:1101: bootstrap] Error 2

  # To reproduce the step with the error, do:
  cd         ~/bigsrc/emacs24/src/
  ./temacs --batch --load loadup bootstrap


I'm not sure of what more I have to send to help people debug this -
please send instructions...

  Cheers =),
    Eduardo Ochs
    http://angg.twu.net/#eev
    http://angg.twu.net/emacsconf2019.html

bug#38868: Problem compiling Emacs24 on Debian 10

[Eli Zaretskii]

> From: Eduardo Ochs <eduardoochs@gmail.com>
> Date: Thu, 2 Jan 2020 03:06:07 -0300
> 
>   git checkout emacs-24
> 
>   ./autogen.sh   2>&1 | tee oa
>   ./configure    2>&1 | tee oc
>   # make         2>&1 | tee om
>   make bootstrap 2>&1 | tee omb
> 
>   # "make bootstrap" fails!
>   tail -n 15 omb
>   # The output of "tail" is:
>   # make[3]: Leaving directory '/home/edrx/bigsrc/emacs24/lisp'
>   # if test "no" = "yes"; then \
>   #   rm -f bootstrap-emacs; \
>   #   ln temacs bootstrap-emacs; \
>   # else \
>   #   ./temacs --batch --load loadup bootstrap || exit 1; \
>   #   test "X" = X ||  -zex emacs; \
>   #   mv -f emacs bootstrap-emacs; \
>   # fi
>   # /bin/bash: line 7: 20593 Segmentation fault      ./temacs --batch
> --load loadup bootstrap

I guess this is some problem with unexec and malloc in new glibc
versions?

bug#38868: Problem compiling Emacs24 on Debian 10

[Stefan Kangas]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: Eduardo Ochs <eduardoochs@gmail.com>
>> Date: Thu, 2 Jan 2020 03:06:07 -0300
>>
>>   git checkout emacs-24
>>
>>   ./autogen.sh   2>&1 | tee oa
>>   ./configure    2>&1 | tee oc
>>   # make         2>&1 | tee om
>>   make bootstrap 2>&1 | tee omb
>>
>>   # "make bootstrap" fails!
>>   tail -n 15 omb
>>   # The output of "tail" is:
>>   # make[3]: Leaving directory '/home/edrx/bigsrc/emacs24/lisp'
>>   # if test "no" = "yes"; then \
>>   #   rm -f bootstrap-emacs; \
>>   #   ln temacs bootstrap-emacs; \
>>   # else \
>>   #   ./temacs --batch --load loadup bootstrap || exit 1; \
>>   #   test "X" = X ||  -zex emacs; \
>>   #   mv -f emacs bootstrap-emacs; \
>>   # fi
>>   # /bin/bash: line 7: 20593 Segmentation fault      ./temacs --batch
>> --load loadup bootstrap
>
> I guess this is some problem with unexec and malloc in new glibc
> versions?

That was 41 weeks ago, and there were no further followups.  It sounds
likely that Eli's analysis here is correct and an older glibc would need
to be used.  In any case, I don't know what we can do about Emacs 24 now.
I'm therefore closing this bug now.

If this conclusion is incorrect and this is still an issue, please reply
to this email (use "Reply to all" in your email client) and we can
reopen the bug report.

bug#38868: Problem compiling Emacs24 on Debian 10

[Basil L. Contovounesios]

Stefan Kangas <stefan@marxist.se> writes:

> Eli Zaretskii <eliz@gnu.org> writes:
>
>>> From: Eduardo Ochs <eduardoochs@gmail.com>
>>> Date: Thu, 2 Jan 2020 03:06:07 -0300
>>>
>>>   git checkout emacs-24
>>>
>>>   ./autogen.sh   2>&1 | tee oa
>>>   ./configure    2>&1 | tee oc
>>>   # make         2>&1 | tee om
>>>   make bootstrap 2>&1 | tee omb
>>>
>>>   # "make bootstrap" fails!
>>>   tail -n 15 omb
>>>   # The output of "tail" is:
>>>   # make[3]: Leaving directory '/home/edrx/bigsrc/emacs24/lisp'
>>>   # if test "no" = "yes"; then \
>>>   #   rm -f bootstrap-emacs; \
>>>   #   ln temacs bootstrap-emacs; \
>>>   # else \
>>>   #   ./temacs --batch --load loadup bootstrap || exit 1; \
>>>   #   test "X" = X ||  -zex emacs; \
>>>   #   mv -f emacs bootstrap-emacs; \
>>>   # fi
>>>   # /bin/bash: line 7: 20593 Segmentation fault      ./temacs --batch
>>> --load loadup bootstrap
>>
>> I guess this is some problem with unexec and malloc in new glibc
>> versions?
>
> That was 41 weeks ago, and there were no further followups.  It sounds
> likely that Eli's analysis here is correct and an older glibc would need
> to be used.  In any case, I don't know what we can do about Emacs 24 now.
> I'm therefore closing this bug now.
>
> If this conclusion is incorrect and this is still an issue, please reply
> to this email (use "Reply to all" in your email client) and we can
> reopen the bug report.

This reminds me of the following report:
https://lists.gnu.org/archive/html/emacs-devel/2019-06/msg01002.html

In which Noam suggests to apply the following patch for building with
-no-pie:
https://sources.debian.org/patches/emacs24/24.5+1-11+deb9u1/0027-Emacs-shouldn-t-segfault-when-gcc-expects-nopie.patch

I recall having to do that on Debian Testing in order to get Emacs 24.5
(from the official tarball, not Debian's) to build.

HTH,

-- 
Basil