[Bug 2076024] Re: SRU: build with -fno-omit-frame-pointer

[Matthias Klose]

** Description changed:

- SRU: build with -fno-omit-frame-pointer, if it's enabled in dpkg-
- buildflags.
+ SRU:
+ 
+   * SRU: LP: #2076024: Build with -fno-omit-frame-pointer.
+   * Build loong64 packages on arm64 and ppc64el.
+   * ld.bfd, ld.gold: When no package-metadata option is given, fall-back
+     to the envvar ELF_PACKAGE_METADATA.
+ 
+ The first one enables the -fno-omit-frame-pointer flag on the
+ architectures where it is the default.
+ 
+ The second one just builds packages which are already built on other
+ architectures on two more architectues.
+ 
+ The third patch is a no-op, because ELF_PACKAGE_DATA is not set by
+ default anywhere in noble.  It gets used by the cross compiler packages
+ to set the correct package name. Without it, cross compiler packages
+ would look like being built from the gcc-N packages, not the gcc-N-cross
+ packages.
+ 
+ These changes were part of a test rebuild of the main component on all architectures, with results at
+ https://people.canonical.com/~ginggs/ftbfs-report/test-rebuild-20240912-noble-tc-noble.html
+ https://people.canonical.com/~ginggs/ftbfs-report/test-rebuild-20240912-noble-noble.html
+ 
+ The first one is built with all proposed changes to the binutils, GCC,
+ LLVM and Python packages, the second one is a reference test rebuild
+ with unmodified release and updates pockets.
+ 
+ Comparing these, we find one progression, and some regressions.
+ 
+ - urwid, progression. builds with the updated packages, fails in the noble
+   archive. Not further investigated.
+ 
+ - linux-* (11 packages). This is LP: #2081797, a configuration issue based
+   on a GCC version.  The fixed compiler already is in the noble release, but
+   still identifies as 13.2.0.  The kernel team is aware of that, and makes
+   the fix of the next upload.
+   A test build with the fix worked fine, and survived a reboot on an amd64
+   laptop.
+ 
+ - gcc-14 fails in the archive, but is part of the planned updates. This is
+   caused by some gnat 64bit time_t fixes.
+ 
+ - bpftrace is a build failure seen with the LLVM 18.1.8 update, however caused
+   by some underlinking in bpftrace. The proposed fix should be SRU'd together
+   with the LLVM update.
+ 
+ - There are some Python related failures, which still need to be investigated.
+   These have to be addressed before the Python SRUs (doing that later than
+   the binutils, GCC and LLVM updates). These are heat, pydantic,
+   python-oslo-config, python-xmlschema and walinuxagent
+ 
+ The packages proposed as SRUs don't show regressions in their testsuites
+ which are run during the builds.

-- 
You received this bug notification because you are a member of Ubuntu
Foundations Bugs, which is subscribed to binutils in Ubuntu.
https://bugs.launchpad.net/bugs/2076024

Title:
  SRU: build with -fno-omit-frame-pointer

Status in binutils package in Ubuntu:
  New

Bug description:
  SRU:

    * SRU: LP: #2076024: Build with -fno-omit-frame-pointer.
    * Build loong64 packages on arm64 and ppc64el.
    * ld.bfd, ld.gold: When no package-metadata option is given, fall-back
      to the envvar ELF_PACKAGE_METADATA.

  The first one enables the -fno-omit-frame-pointer flag on the
  architectures where it is the default.

  The second one just builds packages which are already built on other
  architectures on two more architectues.

  The third patch is a no-op, because ELF_PACKAGE_DATA is not set by
  default anywhere in noble.  It gets used by the cross compiler
  packages to set the correct package name. Without it, cross compiler
  packages would look like being built from the gcc-N packages, not the
  gcc-N-cross packages.

  These changes were part of a test rebuild of the main component on all architectures, with results at
  https://people.canonical.com/~ginggs/ftbfs-report/test-rebuild-20240912-noble-tc-noble.html
  https://people.canonical.com/~ginggs/ftbfs-report/test-rebuild-20240912-noble-noble.html

  The first one is built with all proposed changes to the binutils, GCC,
  LLVM and Python packages, the second one is a reference test rebuild
  with unmodified release and updates pockets.

  Comparing these, we find one progression, and some regressions.

  - urwid, progression. builds with the updated packages, fails in the noble
    archive. Not further investigated.

  - linux-* (11 packages). This is LP: #2081797, a configuration issue based
    on a GCC version.  The fixed compiler already is in the noble release, but
    still identifies as 13.2.0.  The kernel team is aware of that, and makes
    the fix of the next upload.
    A test build with the fix worked fine, and survived a reboot on an amd64
    laptop.

  - gcc-14 fails in the archive, but is part of the planned updates. This is
    caused by some gnat 64bit time_t fixes.

  - bpftrace is a build failure seen with the LLVM 18.1.8 update, however caused
    by some underlinking in bpftrace. The proposed fix should be SRU'd together
    with the LLVM update.

  - There are some Python related failures, which still need to be investigated.
    These have to be addressed before the Python SRUs (doing that later than
    the binutils, GCC and LLVM updates). These are heat, pydantic,
    python-oslo-config, python-xmlschema and walinuxagent

  The packages proposed as SRUs don't show regressions in their
  testsuites which are run during the builds.

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/binutils/+bug/2076024/+subscriptions