erichkeane wrote: > Is there anything we can do to make this change less disruptive going forward? > > Should some Clang experts review the relands? Is there more testing that > could be done? Is there any chance of breaking it into smaller patches?
Maintainers have definitely been discussing this, but unfortunately we haven't come up with a solution. This is a pretty immutable patch (we tried breaking it up, but there isn't really a sane place to do so) as it is a pretty big rewrite of some very interdependent code. Of course, because of the same, a review that finds everything is pretty difficult (AND wild-template uses are... quite varied). It is unfortunate and painful, but we're going to have to keep doing these cycles as long as we can. The best thing we can do is back-feed as much as we can as input to limit the churn. Unfortunately/Ironically, as quickly as we have been reverting actually worsens this, as the revert tends to prevent feedback from being discovered or shared (I had this same problem with a patch a few years ago that significantly changed how concepts works). It IS a bit of a shame we didn't think to try the libcxx tests (and perhaps @mizvekov should make sure to increase his local test breadth with single/multi source tests/etc as well), but I think our only way forward is to keep trying until it sticks. https://github.com/llvm/llvm-project/pull/208285 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
