Bugzilla

I rebased all the dependencies onto the revision in question. With that in place, the behaviour is the same as the behaviour on recent m-c with the fix backed out.

(In reply to Botond Ballo [:botond] from comment #4)
> I tried to back out the fix locally, to better understand how the bug was originally triggered, but I'm finding that I can no longer reproduce the bug (as described in bug 1484597 comment 7) with the fix backed out (on either the reduced testcase or the original).

I suspect that the explanation for this is that the bug described in bug 1484597 comment 7 was fixed by the "Use the visual viewport offset in ScrollToShowRect()" patch in that bug. The "clear the APZ callback transform" patch is fixing a potential issue that I noticed while debugging that bug, but doesn't actually occur with the "Use the visual viewport offset in ScrollToShowRect()" patch in place (but could still occur in other scenarios; in other words, I maintain it's conceptually a valid and important change).

I rebased all the dependencies onto the revision in question. With that in place, the behaviour is the same as the behaviour on recent m-c with the fix backed out.

(In reply to Botond Ballo [:botond] from comment #4)
> I tried to back out the fix locally, to better understand how the bug was originally triggered, but I'm finding that I can no longer reproduce the bug (as described in bug 1484597 comment 7) with the fix backed out (on either the reduced testcase or the original).

I suspect that the explanation for this is that the bug described in bug 1484597 comment 7 was fixed by the "Use the visual viewport offset in ScrollToShowRect()" patch in that bug. The "clear the APZ callback transform" patch is fixing a potential issue that I noticed while debugging that bug, but doesn't actually occur with that STR with the "Use the visual viewport offset in ScrollToShowRect()" patch in place (but could still occur in other scenarios; in other words, I maintain it's conceptually a valid and important change).

I rebased all the dependencies onto the revision in question. With that in place, the behaviour is the same as the behaviour on recent m-c with the fix backed out.

(In reply to Botond Ballo [:botond] from comment #4)
> I tried to back out the fix locally, to better understand how the bug was originally triggered, but I'm finding that I can no longer reproduce the bug (as described in bug 1484597 comment 7) with the fix backed out (on either the reduced testcase or the original).

I suspect that the explanation for this is that the bug described in bug 1484597 comment 7 was fixed by the "Use the visual viewport offset in ScrollToShowRect()" patch in that bug. The "clear the APZ callback transform" patch is fixing a potential issue that I noticed while debugging that bug, but doesn't actually occur with that STR with the "Use the visual viewport offset in ScrollToShowRect()" patch in place (but could still occur in other scenarios; in other words, I maintain it's conceptually a valid and important change).

So I think the challenge here is nailing down STR / test steps that do trigger the potential issue fixed by that patch.