FTBFS [s390x,ppc64el,riscv64]: -Wmaybe-uninitialized

[Andreas Hasenack]

Builds on s390x, ppc64el, riscv64 are failing with a maybe-unitialized error:

gcc -DHAVE_CONFIG_H -I. -I.. -Wdate-time -D_FORTIFY_SOURCE=2 -DLOCALEDIR=\"/usr/share/locale\" -DLIBDIR=\"/lib/riscv64-linux-gnu\" -DTEST -g -O2 -ffile-prefix-map=/<<PKGBUILDDIR>>=. -fstack-protector-strong -Wformat -Werror=format-security -Wno-error=deprecated-declarations -Wall -Wextra -Wno-unused-parameter -Werror -c -o test_passwd-pam_p11.o `test -f 'pam_p11.c' || echo './'`pam_p11.c
base64.c: In function ‘sc_base64_decode’:
base64.c:89:36: error: ‘i’ may be used uninitialized [-Werror=maybe-uninitialized]
   89 | *out++ = i >> s;
      | ~~^~~~
base64.c:79:22: note: ‘i’ was declared here
   79 | unsigned int i;
      | ^
base64.c:94:20: error: ‘skip’ may be used uninitialized [-Werror=maybe-uninitialized]
   94 | in += skip;
      | ^~
base64.c:78:25: note: ‘skip’ was declared here
   78 | int len = 0, r, skip;
      | ^~~~

On amd64 this does not happen.

Initially I thought it was due to ppc64el using -O3. And, in fact, rebuilding it on ppc64 with -O2 doesn't trip this problem, but OTOH building it on amd64 *with* -O3 also does not trigger the problem. Furthermore, s390x and riscv64 are also using -O2 and the build is failing there too.

Previous builds in LP passed with gcc-11, so this looks like it's something new in gcc-12.

[Andreas Hasenack (ahasenack)]

Filed upstream issue at https://github.com/OpenSC/pam_p11/issues/24

[Andreas Hasenack (ahasenack)]

Uploaded

[Launchpad Janitor (janitor)]

This bug was fixed in the package pam-p11 - 0.3.1-1.1ubuntu1

---------------
pam-p11 (0.3.1-1.1ubuntu1) kinetic; urgency=medium

  * d/p/ftbfs-uninitialized-variables.patch: initialize variables to fix
    ftbfs with gcc-12 in some architectures (LP: #1990194)

 -- Andreas Hasenack <email address hidden> Thu, 22 Sep 2022 12:24:51 -0300