vlsi commented on a change in pull request #1375:
[WIP][CALCITE-3140][CALCITE-3141] Failures in slow tests & CI
URL: https://github.com/apache/calcite/pull/1375#discussion_r349508142
##########
File path: .travis.yml
##########
@@ -23,8 +23,6 @@ matrix:
- jdk: openjdk8
- jdk: openjdk11
- jdk: openjdk11
- env:
- - SLOW_TESTS=Y
Review comment:
>Anyways there are not many PRs who receive feedback in less than an hour.
Regular tests complete within 20min, which is well under 1h.
> In terms of resources yes I agree that it is not great but shouldn't we
let GitHub worry about this
This is a chicken-and-egg problem: Travis has a limited set of executors for
all ASF projects, so running long tests does impact other projects.
See
https://lists.apache.org/thread.html/af52e2a3e865c01596d46374e8b294f2740587dbd59d85e132429b6c@%3Cbuilds.apache.org%3E
In the same way, if other projects could reduce the waste, that would
improve "queueing" times for Calcite as well.
Even though some `slow` tests might be important, I suspect that the notable
part of the `slow` tests does nothing but burning CPU (all of them?).
For instance, in
https://github.com/apache/calcite/commit/f2f25b3a567ee7ec1c6e8f9f00f581b2dacc3dae
I "reduced" the heap consumption of `FoodmartTest` by 400MiB. Does that mean
the test was using the memory for a good reason? I doubt so. I think it was
just a bug.
It is clear that launching `slow` tests for each and every commit would burn
a lot of CPU, and the gain would be invisible.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
