Summary: | DocumentLoader::detachFromFrame() is being called with no current Frame set | ||||||
---|---|---|---|---|---|---|---|
Product: | WebKit | Reporter: | Brady Eidson <beidson> | ||||
Component: | Page Loading | Assignee: | Brady Eidson <beidson> | ||||
Status: | RESOLVED FIXED | ||||||
Severity: | Normal | CC: | commit-queue, japhet, joepeck, ossy, thorton | ||||
Priority: | P2 | ||||||
Version: | 528+ (Nightly build) | ||||||
Hardware: | All | ||||||
OS: | All | ||||||
Attachments: |
|
Description
Brady Eidson
2015-07-09 08:42:54 PDT
Created attachment 256498 [details]
Patch v1
I'm still going deeper down the rabbit hole trying to theorize how this might happen, but this patch definitely stops this specific crash.
(In reply to comment #2) > https://trac.webkit.org/changeset/186642 It made 46 tests assert. (In reply to comment #3) > (In reply to comment #2) > > https://trac.webkit.org/changeset/186642 > > It made 46 tests assert. I've said it before, I'll say it again - EWS is useless in its current form :( ACK - DocumentLoaders in the page cache don't detach from their frame =/ Which means when we bring them out and attach to their new frame, they never undergo a formal "detach from frame" process =/ (In reply to comment #5) > ACK - DocumentLoaders in the page cache don't detach from their frame =/ > > Which means when we bring them out and attach to their new frame, they never > undergo a formal "detach from frame" process =/ I see, the clause I removed in the patch covered this - they can only ever come out to the frame they had when going in. This makes me feel a little better, but not super great. (In reply to comment #4) > (In reply to comment #3) > > (In reply to comment #2) > > > https://trac.webkit.org/changeset/186642 > > > > It made 46 tests assert. > > I've said it before, I'll say it again - EWS is useless in its current form > :( I wouldn't say if EWS is useless. But it isn't able to catch assertions yet, because we have only release bots and not debug bots too. You can inspire your company to setup debug EWS bots too. ;) Anyway you can (or should) watch the buildbots after landing. (In reply to comment #7) > (In reply to comment #4) > > (In reply to comment #3) > > > (In reply to comment #2) > > > > https://trac.webkit.org/changeset/186642 > > > > > > It made 46 tests assert. > > > > I've said it before, I'll say it again - EWS is useless in its current form > > :( > > I wouldn't say if EWS is useless. But it isn't able to catch assertions > yet, because we have only release bots and not debug bots too. This is one aspect of the automation paradox. If the automation doesn't do the whole job, then relying on it at all causes problems. > > You can inspire your company to setup debug EWS bots too. ;) My comment was actually a flippant reminder to those who know that I've been asking for 100% EWS coverage for a few years now :) > Anyway you can (or should) watch the buildbots after landing. The other aspect of the automation paradox - Relying on the automation at all causes us to lose our expertise in doing things manually. :) That said, we have bot watchers, non of which who pinged me... yikes! Anyways, a fix is on its way. https://trac.webkit.org/changeset/186677 to restore the frame equality check and resolve the assert. |