RESOLVED FIXED 174028
Force crashWithInfo to be out of line.
https://bugs.webkit.org/show_bug.cgi?id=174028
Summary Force crashWithInfo to be out of line.
Keith Miller
Reported 2017-06-30 10:45:23 PDT
Force crashWithInfo to be out of line.
Attachments
Patch (12.38 KB, patch)
2017-06-30 10:57 PDT, Keith Miller
no flags
Patch (12.38 KB, patch)
2017-06-30 11:02 PDT, Keith Miller
no flags
Patch (13.19 KB, patch)
2017-06-30 13:29 PDT, Keith Miller
no flags
Patch (9.16 KB, patch)
2017-06-30 13:50 PDT, Keith Miller
no flags
Patch (12.39 KB, patch)
2017-06-30 14:09 PDT, Keith Miller
no flags
Patch for landing (12.58 KB, patch)
2017-06-30 14:31 PDT, Keith Miller
no flags
Patch for landing (12.59 KB, patch)
2017-06-30 14:49 PDT, Keith Miller
no flags
Patch for landing (12.59 KB, patch)
2017-06-30 16:56 PDT, Keith Miller
no flags
Patch for landing (12.58 KB, patch)
2017-06-30 17:44 PDT, Keith Miller
no flags
Patch for landing (12.82 KB, patch)
2017-06-30 18:07 PDT, Keith Miller
no flags
Keith Miller
Comment 1 2017-06-30 10:57:48 PDT
Keith Miller
Comment 2 2017-06-30 11:02:54 PDT
Filip Pizlo
Comment 3 2017-06-30 11:07:19 PDT
Comment on attachment 314275 [details] Patch This will still lead to coalescing. You need to feed reason, __FILE__, __LINE__, and probably WTF_PRETTY_FUNCTION for good measure. Remember, any operation of the form op(a, b, c) in tail position will get coalesced with any other op(a, b, c) in tail position. A crash macro is in tail position. I don't think you're passing enough interesting things in __VA_ARGS__ to deduplicate this.
Filip Pizlo
Comment 4 2017-06-30 11:08:27 PDT
Comment on attachment 314276 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=314276&action=review > Source/JavaScriptCore/dfg/DFGGraph.h:101 > - (graph).logAssertionFailure( \ > + (graph).logAssertionFailure( \ > (node), __FILE__, __LINE__, WTF_PRETTY_FUNCTION, #assertion); \ > - CRASH_WITH_INFO(__VA_ARGS__); \ > + CRASH_WITH_SECURITY_IMPLICATION_AND_INFO(__VA_ARGS__); \ This doesn't fix coalescing. Please pass #assertion, __FILE__, __LINE__, and WTF_PRETTY_FUNCTION to CRASH_WITH_SECURITY_IMPLICATION, so that it gets deduplicated appropriately. Ideally, you would have this functionality inside logAssertionFailure, so that you don't have to pass the same things twice.
Keith Miller
Comment 5 2017-06-30 13:29:52 PDT
Keith Miller
Comment 6 2017-06-30 13:50:45 PDT
Keith Miller
Comment 7 2017-06-30 14:09:47 PDT
Keith Miller
Comment 8 2017-06-30 14:12:49 PDT
(In reply to Filip Pizlo from comment #3) > Comment on attachment 314275 [details] > Patch > > This will still lead to coalescing. You need to feed reason, __FILE__, > __LINE__, and probably WTF_PRETTY_FUNCTION for good measure. > > Remember, any operation of the form op(a, b, c) in tail position will get > coalesced with any other op(a, b, c) in tail position. A crash macro is in > tail position. I don't think you're passing enough interesting things in > __VA_ARGS__ to deduplicate this. I ended up passing __FILE__, __LINE__, WTF_PRETTY_FUNCTION, and __COUNTER__. Hopefully, that's enough junk. I wish there was an attribute for this.
Keith Miller
Comment 9 2017-06-30 14:31:24 PDT
Created attachment 314293 [details] Patch for landing
Keith Miller
Comment 10 2017-06-30 14:49:19 PDT
Created attachment 314296 [details] Patch for landing
Keith Miller
Comment 11 2017-06-30 16:56:19 PDT
Created attachment 314325 [details] Patch for landing
Keith Miller
Comment 12 2017-06-30 17:44:05 PDT
Created attachment 314336 [details] Patch for landing
Keith Miller
Comment 13 2017-06-30 18:07:51 PDT
Created attachment 314341 [details] Patch for landing
WebKit Commit Bot
Comment 14 2017-06-30 19:42:50 PDT
The commit-queue encountered the following flaky tests while processing attachment 314341 [details]: fast/workers/worker-exception-during-navigation.html bug 174058 The commit-queue is continuing to process your patch.
WebKit Commit Bot
Comment 15 2017-06-30 19:42:51 PDT
The commit-queue encountered the following flaky tests while processing attachment 314341 [details]: The commit-queue is continuing to process your patch.
WebKit Commit Bot
Comment 16 2017-06-30 23:24:47 PDT
Comment on attachment 314341 [details] Patch for landing Clearing flags on attachment: 314341 Committed r219046: <http://trac.webkit.org/changeset/219046>
WebKit Commit Bot
Comment 17 2017-06-30 23:24:49 PDT
All reviewed patches have been landed. Closing bug.
Keith Miller
Comment 18 2017-07-02 15:17:59 PDT
Note You need to log in before you can comment on or make changes to this bug.