This is an opportunistic merge: there were no conflicts but one, and that one was trivial to resolve ("take both").
A few tests are failing but when I managed to build the current head of 3.6, the tests on Windows were failing as well, so I'm not sure the failures are because of the merge (at least not all of them). Either way, something to be addressed in a separate PR.
I am submitting this merge to make 3.6 buildable again and because I am planning to make a change to master that should not land on 3.6. So to make it easier to isolate that commit.
Note: in order to maintain git commit parents, squash cannot be used for merging.
This is an opportunistic merge: there were no conflicts but one, and that one was trivial to resolve ("take both").
A few tests are failing but when I managed to build the current head of 3.6, the tests on Windows were failing as well, so I'm not sure the failures are because of the merge (at least not all of them). Either way, something to be addressed in a separate PR.
I am submitting this merge to make 3.6 buildable again and because I am planning to make a change to master that should not land on 3.6. So to make it easier to isolate that commit.
Note: in order to maintain git commit parents, squash cannot be used for merging.