Externals Update, master branch (2020.12.09.)
Updated all projects to atlasexternals-2.0.92. The changes wrt. atlasexternals-2.0.91 are the following (atlasexternals@2.0.91...2.0.92):
- Modernised the CMake configuration files exported/installed by
atlas_project(...)
; - Added
Findnumpy.cmake
andFindmatplotlib.cmake
to AtlasLCG (pinging @amete); - Fixed the (standalone) build of XRootD on macOS;
- Added CLHEP and yampl to
AthAnalysisExternals
.
The addition of Findnumpy.cmake
and Findmatplotlib.cmake
came up in !38852 (merged).
The tweaks to AthAnalysisExternals
are necessary for ATLASG-1552 and !39015 (merged).
But the dangerous part of this update is the change made in AtlasCMake (atlasexternals!780 (merged)). It, along with all the other updates, does not require a full-build. However if I made a mistake with it, it has the chance of seriously breaking the nightlies. As discussed with @fwinkl in atlasexternals!780 (merged), I would propose to only merge this in on Friday. (Assuming that the CI will be all happy...) So that any issues with the nightly could be sorted out during the weekend.
Merge request reports
Activity
added RC Attention Required label
added Build master review-pending-level-1 labels
CI Result SUCCESS (hash dfbcf4cb)Athena AthSimulation AthGeneration AnalysisBase 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
AthGeneration: number of compilation errors 0, warnings 0
AnalysisBase: number of compilation errors 0, warnings 0
For experts only: Jenkins output [CI-MERGE-REQUEST-CC7 24891]I'll approve this from the L1 shifter side, but reemphasise that RC Attention Required Tom - L1 shifter
added review-approved label and removed review-pending-level-1 label
mentioned in merge request !39015 (merged)
mentioned in commit afe990f0
added sweep:ignore label
So the update works... fine. But @jburr did find something that I did not expect... (ATLINFR-3895) On top of the latest nightlies, when you "don't recompile enough packages" as part of your partial build, CMake may produce a non-functional build configuration.
As I wrote in ATLINFR-3895 as well, this is just shining light on an issue that we had since forever. As with our previous setup we were still able to build packages incorrectly in such setups. It would just not show up quite so visibly during the build.
So this is just to let a few more people know, that if developers start complaining in the coming days that their
WorkDir
based builds don't work as they would expect them to, it will likely be because they have not set up all the packages for recompilation that they would need to.
mentioned in merge request !39215 (merged)