Externals Update, master branch (2020.05.01.)

Switched all projects to atlasexternals-2.0.65.

The changes wrt. atlasexternals-2.0.64 are (atlasexternals@2.0.64...2.0.65):

• Forced GeoModelCore to be built with optimisations, even in Debug builds (thanks to @ssnyder);
• Updated the AtlasLCG modules to consistently set up <PACKAGE>_<executable>_EXECUTABLE variables for externals that provide executables (in preparation for atlasexternals!664 (closed));
• Updated FindXercesC.cmake to wrap CMake's own FindXercesC.cmake module (in preparation for atlasexternals!664 (closed));
• Updated GPerfTools to use Findgperftools.cmake instead of Findtcmalloc.cmake (in preparation for atlasexternals!664 (closed));
• Updated atlas_build_run.sh to handle quoted commands correctly (in preparation for atlasexternals!664 (closed));
• Fixed the build of Python 2.7 on systems that don't provide Python 2 in /usr (to fix the build of AnalysisBaseExternal on Ubuntu 20.04 );
• Updated GDB to version 9.1 (see ATLINFR-3516 and ATLINFR-3518);
• Command-line usage improvements in flake8_atlas (thanks to @fwinkl);
• Removed the EXTRA_PATTERNS argument from atlas_add_test(...) (thanks to @fwinkl).

Note that I believe this update should succeed in an incremental build as well. But I guess we'll see...

This merge request affects 6 packages:

• Projects/AnalysisBase
• Projects/AthDataQuality
• Projects/AthGeneration
• Projects/AthSimulation
• Projects/Athena
• Projects/VP1Light

Adding @jchapman ,@akraszna ,@krumnack ,@vpascuzz ,@rbianchi as watchers

added Build master review-pending-level-1 labels

CI Result SUCCESS (hash f1edabc9)

	Athena	AthSimulation	AnalysisBase	AthGeneration
externals
cmake
make
required tests
optional tests

Full details available on this CI monitor view
Athena: number of compilation errors 0, warnings 0
AthSimulation: number of compilation errors 0, warnings 0
AnalysisBase: number of compilation errors 0, warnings 0
AthGeneration: number of compilation errors 0, warnings 0
For experts only: Jenkins output [CI-MERGE-REQUEST-CC7 13368]

This looks OK, the CI is clear.

MLB (L1)

added review-approved label and removed review-pending-level-1 label

merged

mentioned in commit 01373744

I'm afraid I didn't test this well enough... In a fresh local build I'm getting all sorts of problems from this update...

The nightly is already running, so not much point in reverting this now, but be prepared for a good number of failures in tonight's nightly related to the Python checks using flake8_atlas...

There are basically 2 problems:

• The update I made for atlas_build_run.sh (atlasexternals@7a893c86) doesn't interact well with the usage of flake8_atlas. "Luckily" this doesn't show up with -DCTEST_USE_LAUNCHERS=TRUE, so it won't show up in the nightly. But local builds on top of the nightly will be affected.
• With flake8_atlas now communicating problems correctly, it's finding a fair amount. In packages that use atlas_install_python_modules(python/*.py), without having any python files in them. As it turns out, we have a good number of those, which were left over from the cmt2cmake.py transition.

The good news is that the latter errors should not break the nightly per se. But they will make a good number of packages be marked as failed. These we shall fix tomorrow.

For the issue with atlas_build_run.sh I'll need to make a fix in atlas/atlasexternals.

For the first issue, I have a possible fix described here: atlasexternals@7a893c86 (comment 3433033)

For the second, I will make a MR now that just removes all obsolete atlas_install_python_modules declarations.

I'm already in the process of making one. In which I want to make some other improvements to the configurations of those packages as well...

Since I think "mentions" do not send out e-mail notifications, my fixes are in !32691 (merged).

After I get a coffee, I'll look at the issue with atlas_build_run.sh...

added sweep:ignore label

mentioned in commit atlasexternals@7a893c86

mentioned in merge request !32691 (merged)

mentioned in merge request !32641 (merged)

mentioned in merge request !32697 (merged)

Hi @fwinkl and @akraszna, I still have to use cmake -DCTEST_USE_LAUNCHERS=TRUE ... for local compilation of "python" packages (e.g. RecExCommon, RecJobTransforms) in master nightly r07 and r08. Is this expected? Thanks.

Yes. Until !32697 (merged) gets merged...