jonkeane opened a new pull request #10345:
URL: https://github.com/apache/arrow/pull/10345
This runs reverse dependency checks using {revdepchecks}. The way that works
is by installing a release version of arrow and the current development version
(i.e. from the git checkout), and then runs checks on each of the reverse
dependencies first with the release (called "old" in {revdepcheck}'s terms) and
with the development version ("new" in {revdepcheck}'s terms). Then it compares
the outputs and will only fail if there is a failure in the new check that is
not in the old check.
I've customized the output a bit so that it prints any errors that come up
in either (in the revdepcheck problems step) so we can more easily diagnose,
but it will only fail if there are new errors.
One thing that I tried and was unable to do is to find a way to cache
packages+info across runs. The github cache action will create a cache, but
because of how they are run on crossbow (i.e. on different branches) the caches
are never accessible in different runs. It looks like Azure has similar
restrictions — is there a different way that we use for caching across crossbow
runs? It's not super critical since this will really be only run when manually
called, but it does take _a while_ to run, which could be sped up quite a bit
if we could save some of the outputs.
--
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]
