[
https://issues.apache.org/jira/browse/TIKA-4820?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18104779#comment-18104779
]
Tim Allison commented on TIKA-4820:
-----------------------------------
The PR above does the following.
Claude estimates PR wall clock should go from ~48 min to ~22, and the PR energy
should go down by 81%.
This cancels superseded PR runs.
This splits pipes/service integration tests out of jdk17 into their own build
so that the two can run in parallel.
This moves the Turkish locale tests to Linux. This is purely a within-jvm test
and should behave the same across platforms. Linux build is faster. Locale test
is now push only and it skips Docker integration tests, which is fine for now.
I think. Let me know if you disagree.
This collapses three Windows jobs into one.
The one change that makes me mildly uncomfortable is that this triggers Windows
only on push, not on a new PR. This means that path issues won't be discovered
until they land on main. I'm ok with this because these are rare, and I don't
see pushing to main and then making a release before the full CI passes.
However, we can change this behavior so that we're running Windows on every PR
as we are now if desired.
> CI: e2e testing and other improvements
> --------------------------------------
>
> Key: TIKA-4820
> URL: https://issues.apache.org/jira/browse/TIKA-4820
> Project: Tika
> Issue Type: Task
> Reporter: Tim Allison
> Priority: Minor
>
> While working on the tika-server refinements, I noticed that e2e was
> accidentally disabled a month or so ago. I turned it back on and got it
> working yesterday.
> The build on ci is now taking 40-50 minutes. Let's see what we can do to
> improve ci time.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)