libpmemobj-cpp upgrade to 1.13.0-2 causes issues (on ppc64el only)

Bug #1943446 reported by Frank Heimes
10
This bug affects 1 person
Affects Status Importance Assigned to Milestone
The Ubuntu-power-systems project
Opinion
Undecided
Ubuntu on IBM Power Systems Bug Triage
libpmemobj-cpp (Ubuntu)
Opinion
Undecided
bugproxy

Bug Description

Canonical engineering recently started to upgrade the universe package “libpmemobj-cpp” to 1.13.0-2 (on impish) and ran into some issues while rebuilding.
And these seem to be ppc64el specific regressions.

Now asking the IBM Power team if this is known and if there are any ideas:

https://launchpad.net/ubuntu/+source/libpmemobj-cpp/1.13.0-2/

https://launchpad.net/ubuntu/+source/libpmemobj-cpp/1.13.0-2/+build/21966756

https://launchpadlibrarian.net/557792526/buildlog_ubuntu-impish-ppc64el.libpmemobj-cpp_1.13.0-2_BUILDING.txt.gz

“obj-powerpc64le-linux-gnu/tests/segment_vector_array_expsize_assign_exceptions_oom /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/test/segment_vector_array_expsize_assign_exceptions_oom_0_none/testfile failed: 139”

Frank Heimes (fheimes)
Changed in libpmemobj-cpp (Ubuntu):
assignee: nobody → bugproxy (bugproxy)
Changed in ubuntu-power-systems:
assignee: nobody → Ubuntu on IBM Power Systems Bug Triage (ubuntu-power-triage)
status: New → Opinion
Changed in libpmemobj-cpp (Ubuntu):
status: New → Opinion
Frank Heimes (fheimes)
summary: - libpmemobj-cpp upgrade to 1.13.0-2 causes issues on ppc64el
+ libpmemobj-cpp upgrade to 1.13.0-2 causes issues (on ppc64el only)
tags: added: universe
description: updated
Revision history for this message
James O'Connor (jpoc) wrote (last edit ):

Same/similar libpmemobj-cpp-1.13.0-2 source seems to have built ok for Debian sid ppc64el

https://buildd.debian.org/status/logs.php?pkg=libpmemobj-cpp

On impish it seems all the make check errors are oom related

The following tests FAILED:
 270 - segment_vector_array_expsize_assign_exceptions_oom_0_none (Failed)
 271 - segment_vector_array_expsize_assign_exceptions_oom_0_memcheck (Failed)
 282 - segment_vector_array_expsize_capacity_exceptions_oom_0_none (Failed)
 283 - segment_vector_array_expsize_capacity_exceptions_oom_0_memcheck (Failed)
 310 - segment_vector_array_expsize_modifiers_exceptions_oom_0_none (Failed)
 311 - segment_vector_array_expsize_modifiers_exceptions_oom_0_memcheck (Failed)
 332 - segment_vector_vector_expsize_assign_exceptions_oom_0_none (Failed)
 333 - segment_vector_vector_expsize_assign_exceptions_oom_0_memcheck (Failed)
 344 - segment_vector_vector_expsize_capacity_exceptions_oom_0_none (Failed)
 345 - segment_vector_vector_expsize_capacity_exceptions_oom_0_memcheck (Failed)
 372 - segment_vector_vector_expsize_modifiers_exceptions_oom_0_none (Failed)
 373 - segment_vector_vector_expsize_modifiers_exceptions_oom_0_memcheck (Failed)
 394 - segment_vector_vector_fixedsize_assign_exceptions_oom_0_none (Failed)
 395 - segment_vector_vector_fixedsize_assign_exceptions_oom_0_memcheck (Failed)

I'll keep digging for differences between the debian and ubuntu build environments.

Revision history for this message
James O'Connor (jpoc) wrote :

From frediz:

@jpoc seems like compiling with no optimization makes the tests pass. I know that one of the differences between Ubuntu and Debian is the level of optimization in the gcc default profile

Also using gcc-10 with optimization works.. so there's something with gcc-11 and optimization flags

Revision history for this message
Frédéric Bonnard (frediz) wrote :

Ok so since gcc-11, it won't pass tests with -O3, neither O2.
Got it work with O1.
```
export DEB_CXXFLAGS_MAINT_STRIP = -O3
export DEB_CXXFLAGS_MAINT_APPEND = -O1
```

As it worked fine as is with gcc-10 on Impish, I guess it's definitely
something for the toolchain team.

Revision history for this message
Frank Heimes (fheimes) wrote (last edit ):

Thx Frédéric and thx James,
look like there are some more cases like this with gcc 11 and higher optimizations enabled:
https://bugs.launchpad.net/ubuntu/+source/mesa/+bug/1943818
https://bugs.launchpad.net/ubuntu/+source/ruby2.7/+bug/1943823
Would be indeed good if the toolchain team can have a look, since I think they are all similar.

Revision history for this message
James O'Connor (jpoc) wrote :

I am still working on isolating this bug and reporting with gcc upstream. To get further buy in from the IBM power toolchain team, I'm reproducing with the IBM Power Advance Toolchain 15.0-1 based on gcc-11.2.1-ee1df65e9233

https://www.ibm.com/support/pages/advtool-install

https://github.com/advancetoolchain/advance-toolchain/releases/tag/at15.0-1

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.