Stefan Sperling wrote on Sat, 28 Mar 2020 11:32 +0100:
> On Fri, Mar 27, 2020 at 10:58:48PM +0000, Daniel Shahaf wrote:
> > This reminds me that we don't have good test coverage for on-disk and
> > over-the-wire interoperability with older versions (other than
> > upgrade_tests_data/, which is wc-only). This would be a good thing to
> > set up some day…
>
> Yes, I totally agree!
>
> It would be great to have better automated test coverage for this.
> Even if it was just one or more buildbots that ran our tests with
> various permutations of supported server/client versions.
Why don't we set this up, then? Your unix-build distro is exactly what we
need to get this off the ground: we should be able to use it to easily
compile and install multiple old versions to different prefixes. We'll
then be able to use «make check URL=…» to test svn 1.A and svnadmin 1.A
against svnserve 1.B or mod_dav_svn 1.B. (I wonder if svnadmin 1.B
shouldn't be in the picture, but we can worry about this later.)
Volunteers, please?
We can prototype the setup on svn-qavm.a.o, if needed, before making it
a proper buildbot. (That'll accelerate the edit-compile-fix cycle.)
> There's a note in our release preparation guidelines that recommends
> that the RM should do some testing like that before rolling a release.
> I think asking RMs to do this manually is asking too much.
+1
This is no small task, and the RM has enough on their plate as it is.
It might be reasonable to ask the RM to _orchestrate_ testing (in order
to ensure that a variety of permutations get covered), but that's not
much more work than setting up a wiki page for recording the permutations
that have been covered.
Cheers,
Daniel
Received on 2020-03-28 22:54:44 CET