Externals Update, 21.0 branch (2020.10.06.)

Updated all projects to atlasexternals-1.0.48.6. The changes wrt. atlasexternals-1.0.48.5 are (atlasexternals@1.0.48.5...1.0.48.6):

• Stopped FindDoxygen.cmake from setting up Pango and Cairo automatically for the runtime environment;
  • This was a cherry-pick from the master and 1.0 branches of atlas/atlasexternals from a while ago;
• Fixed the RPM dependency setup of FindCairo.cmake;
  • I learned it just now that the RPM/package changed its name from Cairo to cairo between LCG_91 and LCG_92...

This is necessary to make the RPMs built for the x86_64-centos7-gcc62-opt platform installable.

Pinging @jchapman.

added bugfix label

This merge request affects 5 packages:

• Projects/AthDataQuality
• Projects/AthDerivation
• Projects/AthSimulation
• Projects/Athena
• Projects/AthenaP1

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

added 21.0 Build review-pending-level-1 labels

Thanks again @akraszna!

added alsoTargeting:21.3 alsoTargeting:21.9 labels

Let's see if we can automate the cherry-picks. (I just thought of doing this...)

Note that we do not need this for 21.0-mc16a, 21.0-mc16d, 21.2 or 21.6.

But 21.3 and 21.9 do need it...

CI Result SUCCESS (hash 78065504)

	Athena	AthDataQuality	AthSimulation
externals
cmake
make
required tests
optional tests

Full details available on this CI monitor view
Athena: number of compilation errors 0, warnings 161
AthDataQuality: number of compilation errors 0, warnings 1
AthSimulation: number of compilation errors 0, warnings 6
For experts only: Jenkins output [CI-MERGE-REQUEST-CC7 44968]

CI passed with many warnings, sending to the L2 for crosscheck.

added review-pending-level-2 label and removed review-pending-level-1 label

Hi @dgil,
The 21.0 nightlies all still contain many warnings - exactly the same number as seen in this merge request in fact:
https://bigpanda.cern.ch/nview/?nightly=21.0_Athena_x86_64-slc6-gcc62-opt
so it's a reasonable bet that the same warnings are appearing here as in the nightly.
Cheers,
John

Hello, I see, thanks. So I think I can approve this one.

added review-approved label and removed review-pending-level-2 label

Could we merge this in for tonight's nightly @rmazini?

merged

mentioned in commit d2ae748a

added sweep:done label

added sweep:failed label

Sweep summary
failed:

• 21.3
• 21.9
• master

It's disappointing that this simple MR was not possible to cherry-pick as-is into 21.3 and 21.9. I'll have to do it myself then...

Hi @akraszna,
I'm happy to quickly make those changes if you want.
Cheers,
John

Please do then. I'm not keen on doing it myself...

I'll start a build on x86_64-centos7-gcc62-opt from 21.0 just not to be sure that these fixes will allow the produced RPMs to really install. (I didn't do a full test previously, I just made sure that the previously "faulty" RPM dependencies would be gone.) But you should open the MRs in the meanwhile. My desktop will be done faster than the CI nodes.

My test build of Athena from the 21.0 branch on the x86_64-centos7-gcc62-opt platform succeeded, and I was able to install the RPMs that I produced. So !37097 (merged) and !37098 (merged) should be good to go into 21.3 and 21.9 respectively. (Once the CI succeeds for them.)

Np - I can do the tasks while sitting in meetings. ;-)

mentioned in commit jchapman/athena@d2c6401c

(The issue is because the GaudiVersions for AthenaP1 and AthDerivations differ from the other projects in 21.0, whereas in 21.3 and 21.9 they all match.)

mentioned in merge request !37097 (merged)

mentioned in commit jchapman/athena@69726ba3

mentioned in merge request !37098 (merged)

removed sweep:failed label

Done in !37097 (merged) for 21.3 and !37098 (merged) for 21.9. Dropping sweep:failed.
Is it worth updating the GaudiVersion for AthenaP1 and AthDerivations in 21.0 to give the automated cherry-picking a helping hand?

Good question. I really hope that once we fix all the remaining issues for the SLC6 and CentOS 7 builds, we would not have to change the externals on the 21.0(...) branches ever again. (For 21.3 and 21.9 that may still be a possibility, but the 21.0 bunch I would really like to leave alone after all of this.)

So I wouldn't bother. That would seem like more work than it would be worth at this point. (Don't forget: A newer/different Gaudi version may also require other changes. I would stay away from that can of worms.)

mentioned in commit c21fd11d

mentioned in commit 222f82de

mentioned in merge request !37305 (merged)

mentioned in commit 5ea3d5dc

mentioned in merge request !37687 (merged)

mentioned in commit a034b886