Comment 18 for bug 1772556

Revision history for this message
Matthew Ruffell (mruffell) wrote :

I looked into the autopkgtest regressions.

The pbuilder and sbuilder failures are due to them attempting to build procenv in a Groovy chroot which was made with debootstrap. procenv currently FTBFS on Groovy, due to changes in GCC10 to -Werror=format-overflow.

```
gcc -DHAVE_CONFIG_H -I. -I.. -I . -I ./platform -I ./platform/linux -D PROCENV_PLATFORM_LINUX -pedantic -std=gnu99 -Wall -Wunused -fstack-protector -Wformat -fdata-sections -ffunction-sections -Werror -g -O2 -MT platform/linux/procenv-platform.o -MD -MP -MF platform/linux/.deps/procenv-platform.Tpo -c -o platform/linux/procenv-platform.o `test -f 'platform/linux/platform.c' || echo './'`platform/linux/platform.c
platform/linux/platform.c: In function ‘handle_proc_branch_linux’:
platform/linux/platform.c:1266:21: error: ‘%s’ directive writing up to 1017 bytes into a region of size 16 [-Werror=format-overflow=]
 1265 | p += 1+strlen ("Name:"); /* jump over tab char */
      | ~~~~~~~~~~~~~~~~~~~~~~~
 1266 | sprintf (name, "%s", p);
      | ^~
In file included from /usr/include/stdio.h:867,
                 from ./platform.h:22,
                 from platform/linux/platform-linux.h:25,
                 from platform/linux/platform.c:19:
/usr/include/x86_64-linux-gnu/bits/stdio2.h:36:10: note: ‘__builtin___sprintf_chk’ output between 1 and 1018 bytes into a destination of size 16
   36 | return __builtin___sprintf_chk (__s, __USE_FORTIFY_LEVEL - 1,
      | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
   37 | __bos (__s), __fmt, __va_arg_pack ());
      | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
platform/linux/platform.c:1271:21: error: ‘%s’ directive writing up to 1017 bytes into a region of size 16 [-Werror=format-overflow=]
 1270 | p += 1+strlen ("PPid:"); /* jump over tab char */
      | ~~~~~~~~~~~~~~~~~~~~~~~
 1271 | sprintf (ppid, "%s", p);
      | ^~
In file included from /usr/include/stdio.h:867,
                 from ./platform.h:22,
                 from platform/linux/platform-linux.h:25,
                 from platform/linux/platform.c:19:
/usr/include/x86_64-linux-gnu/bits/stdio2.h:36:10: note: ‘__builtin___sprintf_chk’ output between 1 and 1018 bytes into a destination of size 16
   36 | return __builtin___sprintf_chk (__s, __USE_FORTIFY_LEVEL - 1,
      | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
   37 | __bos (__s), __fmt, __va_arg_pack ());
      | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

```

This has nothing to do with the changes made to debootstrap.

Looking at the changelog for procenv, this has happened before...

procenv (0.50-1ubuntu2) eoan; urgency=medium

  * Hack it so it ignores broken gcc overflow warnings on ppc64el
  * Stop mv README in debian/rules, that breaks building twice (it's a symlink already)

 -- Julian Andres Klode <email address hidden> Mon, 05 Aug 2019 13:43:53 +0200

procenv (0.50-1ubuntu1) eoan; urgency=medium

  * debian/patches/gcc-9-strncpy.patch: Handle strncpy in a way
    compatible with gcc-9 linting.

 -- Steve Langasek <email address hidden> Sat, 20 Jul 2019 17:30:12 +0000

Regardless, I have reported the procenv FTBFS upstream on this issue:

https://github.com/jamesodhunt/procenv/issues/15

I have also opened a LP bug 1889138 for Groovy:

https://bugs.launchpad.net/ubuntu/+source/procenv/+bug/1889138

As for the livecd-rootfs autopkgtest regression, I am still investigating.

The errors are:

snap: found lxd=4.0/stable/ubuntu-20.04
Unexpected seeded snap for ubuntu-cpc:minimized build: lxd=4.0/stable/ubuntu-20.04
autopkgtest [18:33:53]: test minimized: - - - - - - - - - - results - - - - - - - - - -
default-bootstraps PASS
minimized FAIL non-zero exit status 1