CASA Build and Test Group Meeting
Tuesday, 11th August, 2015, 11:00; Room ER331.
Mark, Darrell, Ville, Andy, Akeem
Post-meeting edits and additions in blue text.
- Status of current CASA packages
- Release - 4.4.0 out for OS X (10.9 and 10.10) and RHEL 5 and 6
- Pipeline release (VLA-focused) - 4.3.1 out for RHEL 5 and 6; OS X packages available, still to be publicly announced <- VLA and ALMA version subsequently pushed out.
- 4.5 test/stable packages routinely being produced for RHEL5 and 6 <- widebandmosaic test still keeps failing on OS X - we really would like to get this one fixed, as it prevents stable package production without manual intervention. If the test is good then the code needs to be fixed, if the test is failing falsely then the test needs to be fixed (and could possibly be disabled in the interim). Also, some recent failures of spxfit seen.
- Pre-release packages to be produced for RHEL 5, 6 and OS X 10.9 and 10.10, following code freeze (target date: August 17th)
- "Extended" tests being run on the tarball versions. Dirk's M100 test routinely e-mails Mark and Andy. "Extended -> "Accepted" tests. Take about 7 hours to run on RHEL. Pipeline and NGC3256 tests failing. Akeem will nag Dirk P. about these. MPI and some SD tests to be added to the nightly runs.
- Current timeline plan for 4.5:
- Feature freeze deadline passed (start of August). Remaining 4.5 planning timeline (factoring in hard October 1st deadline for ALMA):
- Main feature testing to be done in the next two weeks or so.
- (Pre-)release branch creation scheduled for August 17th.
- GUI testing, etc. to ramp up in second half of August.
- First release candidate available for random user testing: August 31st. GUI spot-checks.
- Final packaging: early September. Current nominal target release date: September 9th.
- We have 10% contributions from Masaya Kuniyoshi and Renaud Miel. Effectively, their contributions are probably primarily libsakura- and NAOJ-deployment related right now. Best way to use their time allocations? Keep them in reserve, or try to get a block of (say) 2.5 days every 5 weeks? <- Noted here that we need to come up with better ways of communicating with our counterparts in Japan.
- Build and Test Review Status
- Casacore unification - Being worked on now. For reference, Jim's proposed plan was as follows:
- 0.5. Attempt to build against the GC casacore and identify problems. Done.
- 0.75. Iteratively fix bugs and attempt to build again. This was Shinnosuke's B&T primary assignment. Done.
- 1. Merge our Casacore into GC casacore. This will require us to flush all casacore mods in and then let Ger do the final merge (finally). Done.
- 2. Build CASA against GC casacore (platforms?) Done.
- 3. Test CASA/GC casacore against our various regression tests, etc. Done.
- 4. Create a Jenkins job that will handle B&T of codebase using GC casacore. Previously discussed.
- 5. Keep a cached, read-only copy of GC casacore locally(?) to prevent a glitch at googlecode/github from allowing us to build. Also we should maintain periodic a backup copies.
- 5.5. Survive the migration from GC to GitHub... Done.
- 6. Go live and live happily ever after. Still a work in progress...
- This process has now been completed, and will be removed from future agendas. Looking forward, the current plan will be to incorporate the external GitHub CasaCore repository as a submodule in the CASA development team Git version control system.
- B&T news/items from EA.
- B&T news/items from Charlottesville.
- Regression testing of OS X (10.10 and 10.9 test packages). Tests mostly seem to be passing at the moment. Would be good to get the widebandmosaic test passing on OS X as well prior to the 4.5 release (ideally, before the code freeze date...)
- Status of write-up / script / whatever of how to build CASA on OS X and Linux (for developers)? Ville? <- The relevant wiki page is pretty much up to date (although the RHEL5 environment details may need the odd minor update).
- Git migration and Continuous Integration Planning
- First full(-ish) version of document produced.
- Mark R. currently on the hook for a 30-minute presentation to the (Charlottesville-based) CASA Users Committee meeting in October. Will need to have a clearer sense of how things will be done in practice (commits to casacore, etc.). Also to talk about Continuous Integration as an idea, and try to get a sense of how the community would feel about the release schedule implications if such an approach were to eventually be adopted.
- GitLab vs. Stash vs. GitHub Enterprise, etc. Some momentum from management for an Atlassian solution (possibly pending their responses to questions previously raised). We would like to be able to run OS X testing on a Linux server (e.g. Travis CI would support this).
- If we do end up with out own NRAO-hosted Git set-up, could we bolt Travis CI onto it and test it on there for a small number of tests? Not insignificant work would be required for this. Darrell and Ville were looking into OpenID as a possible authentication solution that would work within the NRAO IT security framework. The OpenID idea is now effectively something of a moot point, so will not be prusued further.
- Is this a good place to discuss the details of section 11.1 "Automated Build and Testing Support" of the git migration plan? Just a recap on what is going on with this.
- Putting in some logic in the test suite for the system to ignore those tests that will fail, based on test failures earlier in the run. Test failure logic: Maybe run the unit tests first, and if they fail, then don't bother running subsequent longer tests that would consequently fail as well. This would help to keep the test run times down. We could also set up the system to share out tests between (a flexible number) of multiple hosts to keep the run times down. Ideally, doing both of the above would be best.
- CASA test plan - Andy
- As part of the CASA test plan, Jeff is really interested in the automated test schedule.
- Review the test schedule of those machines doing automated testing.
- I gave the numbers to Joe a number of months ago, so I'd like to review what I've got, to make sure it is up to date
- Some discussion took place about test systems and scheduling. We have rh5tst and rh6tst, run at 04:30 EST ("Critical" tests). 2 Macs, razor (10.9) and hybrid (10.10), running at 01:30 EST. The Macs run through the tests slightly more quickly. The machine called warp performs builds from source (casacore from GitHub, everything else comes from SVN) and runs the smoke tests. It checks for changes every 3-5 minutes. All of the inputs are effectively treated as SVN repositories.
- Casaguide update on G192 from Akeem <- Test helpers needed for this.
- Incorporating graphs in to the existing report framework <- Times could be obtained for these by parsing the logs. Some discussion here about where to put the plots. Directly under the test summaries generated by Ville? The plots currently span the last 28 days, but should cover longer. Action Item on Akeem to generate examples that span a 6-month period.
- Getting accepted tests running on OSX. Darrell agreed to help Akeem get the smoke tests running under Gradle on the Macs.
- Review short-term task priority lists
- Mac hardware availability for developers and user testers: still an issue?
- Old Action Items
- Andy: Discuss Jenkins test for check-ins on OS X with Alexis: same (or different?) smoke tests for OS X as for Linux? Related: machines in Socorro for this? (See above). It is noted that this AI is not a pressing issue at this point. Andy had started to have some discussion with Alexis about this. Smoke tests are being run on the new Test Group Macs. They still need to discuss test coverage for OS X and Linux. Ongoing, with the understanding that this is a lower priority item than Alexis' other work, and that his remaining contract period is limited.
- Next meeting: TBA
- Any Other Business
- Mark will be out of the office for (most of) the next two weeks.
New Action Items Arising