|PresShell::AppendReflowCommand| has an early return when there hasn't been the initial reflow yet. Further down the code, any reflow command that isn't added to the queue is deleted. The question is: why doesn't this apply to the first early return?!? === Another separate thought: how common are reflow commands rejected? If these are too common, maybe the signature of AppendReflowCommand could be changed so that commands are not created just to be deleted immediately. This means doing the creation inside the function itself. Of course, it is not worth the hassle if it is rare.

> The question is: why doesn't this apply to the first early return?!? That's a bug, indeed. > maybe the signature of AppendReflowCommand could be changed I think it's worth doing this anyway.... There's no reason to burden every single caller with having to call NS_NewReflowCommand, check the rv, etc. David, if you don't have any objections, I'll go ahead and do just that.

This moves the reflow command creation into presshell and cleans up a bunch of the code. I went ahead and removed some unused members of the reflow command (some were write-only, and some were always null and never read).

Comment on attachment 171331 [details] [diff] [review] Patch r=rbs Also update the description of |GetChildListName()|, as what is left there is outdated and makes little sense.

Comment on attachment 171331 [details] [diff] [review] Patch rbs, would you mind doing r+sr? David said that's OK with him.

Comment on attachment 171331 [details] [diff] [review] Patch sr=rbs

Fixed. Note that I didn't check in quite the patch attached to this bug, because I had to merge to the prescontext arg removal for Append/Remove/ReplaceChild.