Hi
Ed,
Here
is the process we follow,
There
are two steps in promotion.
- Promote build to
downloads pages. To serve users who want to download.
- Add to I-builds
composite repository
When
we find problems with build we mark a build as unstable.
This removes offending build from I-builds composite
repository. This enables end users to use a good build for
upgrade process. These broken builds stay for a week on the
download page and after they are automatically removed. This
process is there for developers to analyse problems.
Due
to this process whenever downstream projects point to
I-builds composite most often they’ll see a good build. We
also mark a build unstable if we find any problems during
the build process itself.
Because
of above mentioned process Tycho will not pull a bad build
in maximum cases.
Hope
this explanation helps
Thanks
Sravan
Hi
There is something much stranger gong on.
I ran an early OCL build that would have run on Sunday. It
has SWT usage but the build runs fine, because it used
https://download.eclipse.org/eclipse/updates/4.20-I-builds/I20210413-1400
although that is the earliest of the 3 listed in the
compositeArtifacts.jar. If it had used the more recent then
the problem might have shown up.
?? Why did Tycho fail to pull in the broken build ??
http://download.eclipse.org/eclipse/downloads/drops4/I20210413-1400/
has 63 test fails which is rather disturbing.
http://download.eclipse.org/eclipse/downloads/drops4/I20210415-0010/,
which is not marked as unstable, has 7057 fails.
Surely such a high number of fails should be a private matter
to be resolved by the platform committers with no more trouble
to the community as a whole than perhaps a polite message to
cross-project-dev announcing the lack of build promotions
while a problem is investigated.
Why does http://download.eclipse.org/eclipse/downloads/
show unstable builds? Surely these too should not have been
promoted to waste disc space / confuse the community?
Why do any builds have any fails at all? Surely a test that
fails is a test that fails? In my code, if a test is proving
problematic then I raise a Bugzilla and comment out the test
until a solution is found.
Regards
Ed Willink
On 16/04/2021 07:45, Niraj Modi wrote:
_______________________________________________
cross-project-issues-dev mailing list
cross-project-issues-dev@xxxxxxxxxxx
To unsubscribe from this list, visit https://www.eclipse.org/mailman/listinfo/cross-project-issues-dev