Problem building ROOT with gcc 11, Ubuntu 22.04 on GitHub Actions

[KrisThielemans]

Check duplicate issues.

• Checked for duplicates

Description

Since a few days, our GitHub Actions are failing when building ROOT via CMake. This is part of a SuperBuild with lots of other projects, and uses a parallel build. The problem is reproducible, i.e. all jobs fail at the same point. However, the failure depends on the ROOT version.

When compiling v6-26-10, we get SyneRBI/SIRF-SuperBuild#857

CMake Error at /home/runner/work/SIRF-SuperBuild/SIRF-SuperBuild/build/builds/ROOT/build/TBB-prefix/src/TBB-stamp/TBB-build-Release.cmake:49 (message):
  Command failed: 2

   'make' 'compiler=gcc' 'cpp0x=1' 'CXXFLAGS= -mno-rtm' 'CPLUS=/usr/lib/ccache/g++-11' 'CONLY=/usr/lib/ccache/gcc-11' 'LDFLAGS='
  See also
    /home/runner/work/SIRF-SuperBuild/SIRF-SuperBuild/build/builds/ROOT/build/TBB-prefix/src/TBB-stamp/TBB-build-*.log

the log file says

build/common.inc:48: "Warning: deprecated cpp0x=1 is used, stdver must be used instead. Building in stdver=c++0x mode."
../../build/Makefile.tbb:28: CONFIG: cfg=debug arch=intel64 compiler=gcc target=linux runtime=cc11.4.0_libc2.35_kernel6.2.0
make[7]: *** read jobs pipe: Bad file descriptor.  Stop.
make[7]: *** Waiting for unfinished jobs....
In file included from ../../src/tbb/concurrent_hash_map.cpp:17:
../../include/tbb/concurrent_hash_map.h: In constructor ‘tbb::interface5::internal::hash_map_base::hash_map_base()’:
../../include/tbb/concurrent_hash_map.h:131:24: warning: ‘void* memset(void*, int, size_t)’ clearing an object of type ‘struct tbb::interface5::internal::hash_map_base::bucket’ with no trivial copy-assignment; use value-initialization instead [-Wclass-memaccess]
  131 |             std::memset(my_embedded_segment, 0, sizeof(my_embedded_segment));
      |             ~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
../../include/tbb/concurrent_hash_map.h:93:16: note: ‘struct tbb::interface5::internal::hash_map_base::bucket’ declared here
   93 |         struct bucket : tbb::internal::no_copy {
      |                ^~~~~~
make[6]: *** [Makefile:28: tbb] Error 2

Trying to upgrade to v6-28-12, https://github.com/SyneRBI/SIRF-SuperBuild/pull/859/files, it compiled TBB ok, but failed in XROOTD installation. XROOTD-install-err.log says

[I] No git repository info found. Trying to interpret VERSION_INFO
[I] The XrdVersion.hh file already exists
gmake[7]: *** read jobs pipe: Bad file descriptor.  Stop.
gmake[7]: *** Waiting for unfinished jobs....
gmake[6]: *** [Makefile:136: all] Error 2
CMake Error at /home/runner/work/SIRF-SuperBuild/SIRF-SuperBuild/build/builds/ROOT/build/builtins/xrootd/XROOTD-prefix/src/XROOTD-stamp/XROOTD-install-Release-impl.cmake:9 (message):
  Command failed (2):

   '/usr/local/bin/cmake' '--build' '.' '--target' 'install'

The same error occurs with v6-30-04.

The log files actually don't really give an error, but the bad file descriptor is strange of course. I'm therefore not sure if this is a problem in ROOT, or a temporary problem on GitHub Actions, or of course a problem with how we build (although that seems to be finally CMake, https://github.com/SyneRBI/SIRF-SuperBuild/blob/master/SuperBuild/External_ROOT.cmake).

Any ideas?

Reproducer

I cannot reproduce this on my own Ubuntu 22.04 machine at present. We do get the failures in our GitHub Actions jobs.

ROOT version

see above

Installation method

CMake

Operating system

Ubuntu 22.04

Additional context

No response

[hahnjo]

@KrisThielemans can you share the full TBB and XRootD error files? The snippets above only show the last lines after make gave up, not what the actual error is (in case of TBB it's only a warning that is not fatal)

[KrisThielemans]

That's all there is, which is confusing! here are 2 zip files with all the log files in the builds
ROOT_v6-28-12-XROOT-error.zip
ROOT-v6-26-10-TBB-error.zip

[hahnjo]

Right, very confusing... The only hint is read jobs pipe: Bad file descriptor. - maybe this means you are running out of file descriptors in excessive parallel builds / on the "restricted" GitHub Actions runners?

[KrisThielemans]

just using 2. https://github.com/SyneRBI/SIRF-SuperBuild/blob/929fa6052d14d2c5f99d8917f5cd0b27c13bece8/.github/workflows/c-cpp.yml#L152

[KrisThielemans]

This has disappeared. Presumably a temporary GHA problem. sorry for the noise.

[KrisThielemans]

FYI, we have the same problem again since about 1 month. I tried using a non-parallel build (SyneRBI/SIRF-SuperBuild#903), which does work, but results in prohibitive build times.

[hahnjo]

Without a reproducer outside for GitHub Actions there is nothing we can realistically do here...

[dpiparo]

@KrisThielemans thanks a lot for reporting. I agree with @hahnjo . We did not see that error on any of our platforms unfortunately.