SocketTimeoutException after migrating from tycho-eclipserun-plugin to tycho-eclipse-plugin

[ewillink]

https://gitlab.eclipse.org/eclipsefdn/helpdesk/-/issues/6715 describes problems with a JustJ promotion job after migrating from tycho-eclipserun-plugin to tycho-eclipse-plugin. It would appear that the handling of flaky internet connections by tycho-eclipse-plugin is inferior to tycho-eclipserun-plugin. Perhaps tycho-eclipse-plugin needs a better retryu default / algorithm.

[laeubi]

I assume you are talking about https://tycho.eclipseprojects.io/doc/latest/tycho-eclipse-plugin/eclipse-run-mojo.html that was moved but not otherwise enhanced.

The mojo just runs an Eclipse application it has no way to know about what it runs or about application specific parameters (like socket timeouts). So it should simply work as before...

In any case you would need to provide more details and likely a PR for what you want to improve (don't forget to include an integration-test to demonstrate the issue)

[ewillink]

Providing an integration test for a flaky internet is something of a challenge.

The problem is demonstrated by: https://ci.eclipse.org/modisco/job/justj-promoter/129/ and fixed in https://ci.eclipse.org/modisco/job/justj-promoter/131/.

The correlation of eclipserun migration and Jenkins failure may be just coincidence. ci.eclipse.org may have degraded instead.

If the mojo itself is unchanged, then it could be that the "tycho.extras" level provided better system defaults.

For some reason adding

<argLine>
-Dorg.eclipse.equinox.p2.transport.ecf.retry=100
</argLine>

is 100% necessary with tycho-eclipse-plugin whereas perhaps 5% necessary with tycho-eclipserun-plugin.

If the above is enough to stimulate an aha, great. If not WONTFIX. I cannot help further than reporting the phenomenon and workaround.