Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [jdt-dev] Almost there! But caution! - ENTERING NEXT LEVEL

Just so that everybody is aware:

Until recently, the job eclipse.jdt.core-Gerrit-BETA stopped just after finishing the compiler test suite, due to failures.

Then Jay fixed/disabled remaining failures to that the build gets passed the first test plugin (which then I broke and refixed :) ).

Point to note: you should now check each Gerrit-BETA build for regressions.
If core.tests.compiler fails, the change is no good, and will again mask any failures in subsequent test plugins.

core.tests.model still has 11 failures (3 in CompletionTest12 and 8 in JavaSearchTest). Once those are fixed, we can hopefully go back to requiring a green build before every push, also in BETA_JAVA_12

BTW: Do we have a collector bug for any tests that were disabled during this exercise?

Other JDT components: do we need additional BETA_JAVA_12 gerrit builds??
Mmh, I don't even know, if JDT/UI and /Debug can see the J12 branch from /Core??

cheers,
Stephan

On 15.01.19 13:57, Stephan Herrmann wrote:
Dear team,

When looking at the test results of releases 4.8 - 4.10, JDT was accumulating a lot of technical debt = failures, some intermittent, some systematic.

As of I20190114-1800 results look much better, as the attached picture demonstrates, the remaining failures being:
1. ClassNameTests.testBug152841
    I can't recall seeing this ever before, only one platform
2. TypeHierarchyNotificationTests.testAddAnonymousInRegion
    intermittent (infrequent)
3. LeakTestSuite complete failure (15) on MacOS

I personally wouldn't pay much attention to (1), (2) is in one group with a few others that happened more than once but too rarely to badly worry about, IMHO.

With (3) Kalyan has a quite hard nut to crack. Let's wish him luck (or help him).

Other than that, I think the current state is good enough for a little toast to the team and particularly to Andrey(!), who has invested a lot recently to get closer to the big zero.

Cheers!


One reason for sending this mail: encourage *every one* in the team to try to keep this status, by monitoring build results below
    https://download.eclipse.org/eclipse/downloads/
and by reacting to any failures. Just watching your own gerrit job doesn't seem to suffice when many bugs only occur on specific machines / os / environment.
(Did Sasi provide his build-observing App to anybody?)
Oh, what should actually be the compliance level of JDT/Core gerrit tests?
Currently it's 10, but perhaps it should already be 11? What pace of update should the job have these days? Can we make this a part of the update-to-next-Java-version routine?

But, the next wave of technical debt could soon be upon us: our BETA gerrit job at https://ci.eclipse.org/jdt/job/eclipse.jdt.core-Gerrit-BETA/ isn't really in a very good shape with constantly 59 failures! Those have been captured in bug
https://bugs.eclipse.org/540704 & children.

 From these, one bug specifically worries me:
https://bugs.eclipse.org/540922 - [12] ct.sym file has been restructured again for older releases

My latest comments are meant to indicate a severe bug in the JEP 247 process: we need to implement the "-release" option but the necessary information is only available in undocumented and changing formats. At the same time people expect JDT to compile against *future* JREs. This set of requirements is impossible to fulfill. Please join me in discussing the options we have (if any) in the bug.

Please also help to eliminate the other failures in BETA_JAVA_12. It is so much easier and nicer to fix individual failures when they occur rather than letting them pile up until we are overwhelmed by the mess. :)

At this point I hardly dare mentioning the back-to-back tests with javac:
   https://ci.eclipse.org/jdt/job/eclipse.jdt.core-run.javac-1.8/
   https://ci.eclipse.org/jdt/job/eclipse.jdt.core-run.javac-10/
which linger at > 200 differences between compilers, but I am monitoring those and didn't see many regressions over time, good.
The last job:
   https://ci.eclipse.org/jdt/job/eclipse.jdt.core-run.javac-11
doesn't even run regularly, I'm re-checking its status right now
(recently it showed some nice crashes of javac ;p ).


Sorry, if this mail sounds a bit negative. Generally I'm very happy to see the recent progress.

best,
Stephan

_______________________________________________
jdt-dev mailing list
jdt-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://www.eclipse.org/mailman/listinfo/jdt-dev




Back to the top