mozilla :: #apz

11 Feb 2017
00:08botondtnikkel: ping?
00:30tnikkelbotond: hey
00:30botondtnikkel: oh hey, good timing! i *just* got a stack of interest to you
00:31botondtnikkel: https://pastebin.mozilla.org/8978749
00:31botondtnikkel: here's NeedsCharsetSwitchTo() calling nsDocumentViewer::LoadComplete()
00:40tnikkelbotond: thats on the first document viewer?
00:43botond|laptoptnikkel: yeah
00:48tnikkelbotond: so doesnt that document viewer get destroyed and a new one created? so that "load complete" state should disappear and in the new presshell it shouldn't think the document is loaded, no?
00:50botondtnikkel: right, the new pres shell doesn't think that. but, in between the old one getting destroyed and the new one getting created, we get into here: http://searchfox.org/mozilla-central/source/layout/generic/nsGfxScrollFrame.cpp#4228, the PageIsStillLoading() call there returns false, and we clear the mRestorePos (http://searchfox.org/mozilla-central/source/layout/generic/nsGfxScrollFrame.cpp#4241)
00:54tnikkelbotond: oh, and then we save that restore pos (or lack thereof) and the new presshell reads that in preference to scrolling to the anchor?
00:54botondtnikkel: yup, exactly!
01:02tnikkelhmmm
01:16tnikkelbotond: this is a wierd case. the only thing i can think of right now is that we need a signal from the parser that it is doing an immediate reload of the page, so don't trust the "page is finished loading" state
01:35tnikkelbotond: wait, do we get a Stop() call on the document viewer? i think we do. that means mStopped is true on the document viewer
01:35tnikkelbotond: that signals that not all the content is loaded
01:35tnikkelwhich is the thing we care about
01:35tnikkelso maybe we can keep trying to restore in that case
01:36tnikkelbut i'd have to look over where and why we want to stop trying to restore to make sure it doesn't regress anything
01:38botond|hometnikkel: ok, i'll look into that. thanks!
12 Feb 2017
No messages
   
Last message: 73 days and 23 hours ago