There was some enthusiasm upstream for the part of my patch (patch 1) that reorders the passes, and I need to get around to testing/benchmark that to make sure it has no undesirable side-effects. The other part of that patch was not approved.
So reviewers upstream expressed a preference for Chung-Lin's patch (patch 2, above), not least because it appears to catch another case mine does not. However, this patch had some bugs that would need fixing before it could be committed.
This is still stuck.
There was some enthusiasm upstream for the part of my patch (patch 1) that reorders the passes, and I need to get around to testing/benchmark that to make sure it has no undesirable side-effects. The other part of that patch was not approved.
So reviewers upstream expressed a preference for Chung-Lin's patch (patch 2, above), not least because it appears to catch another case mine does not. However, this patch had some bugs that would need fixing before it could be committed.