Julian Foad wrote on Tue, 20 Apr 2021 13:58 +00:00: > Julian Foad wrote on 2021-04-14: > > If we are not looking at back-porting it to 1.14.x, then we could > > [...] > > - use the checks in libsvn_client; add a new 'notification' type for > > the warnings > > That way was easiest and didn't require much boiler-plate code at all. > Done in http://svn.apache.org/r1889012 . > > As explained in the comment in https://subversion.apache.org/issue/4874 > this changes the 1.14 behaviour to just add a warning in 1.15 and then > become an error in 1.16. > > For 1.16 a small modification to the test expectations will need to be > made, as I didn't find a simple way to make the expectation depend on > the client version being tested. This will come up as a failure with an > obvious fix, when we bump the trunk version to 1.16 after branching > 1.15. If someone spots a way to automate this part too, do say.
There's actually an svntest.main.SVN_VER_MINOR constant to test. It's not parsed from svn_version.h, but gets updated manually (in https://subversion.apache.org/docs/community-guide/releasing.html#creating-branch; see under "Manual Procedure", fourth bullet), but we can use it all the same. I'll leave it to you to decide whether to test the constant directly or to encapsulate the check in a named has_foo() helper function, as done for other similar checks. (Aside: those <h4/>'s don't have the self-link pilcrows that all other <h{1..4}/>'s do.) > This appears to complete the issue. Do say if I have missed something. Haven't looked at r1889012 yet, but all the same, thanks for your usual thoroughness on this. Cheers, Daniel > -- > - Julian >
