[Bug 1798700] Re: ftbfs on arm64

[Michael Hudson-Doyle]

As above, the build passing is a sufficient test for this.

** Tags removed: verification-needed verification-needed-cosmic
** Tags added: verification-done-cosmic

-- 
You received this bug notification because you are a member of Ubuntu
Foundations Bugs, which is subscribed to elfutils in Ubuntu.
https://bugs.launchpad.net/bugs/1798700

Title:
  ftbfs on arm64

Status in elfutils package in Ubuntu:
  Fix Committed
Status in elfutils source package in Cosmic:
  Fix Committed

Bug description:
  [impact]
  elfutils ftbfs on sid on arm64 currently:

  FAIL: run-backtrace-dwarf.sh
  ============================

  0xffffb3c5373c	raise
  0xffffb3c418e8	abort
  /<<PKGBUILDDIR>>/tests/backtrace-dwarf: dwfl_thread_getframes: (null)
  dwarf: no main
  FAIL run-backtrace-dwarf.sh (exit status: 1)

  FAIL: run-deleted.sh
  ====================

  PID 19405 - process
  TID 19405:
  #0  0x0000ffff81e7f3e0 __nanosleep
  #1  0x0000ffff81e7f2c0 sleep
  /<<PKGBUILDDIR>>/src/stack: dwfl_thread_getframes tid 19405 at 0xffff81e7f2bf in /lib/aarch64-linux-gnu/libc-2.28.so: (null)
  FAIL run-deleted.sh (exit status: 1)

  Current git passes, and git bisect and logic point to the following
  upstream commit as fixing it:

  commit f881459ffc95b6fad51aa055a158ee14814073aa
  Author: Mark Wielaard <mark at klomp.org>
  Date:   Wed Apr 11 10:37:45 2018 +0200

      aarch64: Add default cfi rule to restore SP from CFA address.

      The CFA is set by default to the stack pointer of the previous frame.
      So that is also how we can always restore the SP. This default aarch64
      CFI rule is necessary on Fedora 28 with GCC8 to make the run-deleted.sh
      and run-backtrace-dwarf.sh testcases work.

      Signed-off-by: Mark Wielaard <mark at klomp.org>

  Debugging the test failure made it clear that the libdw1 in the
  archive will not be able to successfully walk the stack of processes
  built by gcc-8, i.e. this is a regression in runtime functionality. I
  don't know if any of the revdeps of libdw1 depend on this but it seems
  at least possible and the patch to fix all this is pretty simple so my
  inclination is that this meets the threshold for an SRU.

  [test case]
  Build the package on arm64.

  [regression potential]
  The patch is small, fairly (given the general level of hairiness around here) simple and only affects the functionality that is currently broken. There's always the chance that building with a new toolchain will introduce new problems, but there's not much we can do about that.

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/elfutils/+bug/1798700/+subscriptions