Detect when Chrome crashes after its bundle has been moved or deleted |
||||||
Issue descriptionOn macOS, moving/deleting Chrome.app or the directory in Contents/Versions that's currently in use will lead to new renderers not starting, and a crash the next time a window is opened. This mainly happens because NSBundle doesn't handle moves: The next time an NSViewController tries to load a nib, it gets the old path, fails, and creates a plain NSView instead (and doesn't hook up any of the view controller's outlets). Several different crashes (685985, 526276, 528136) all look like the result of this. The crashes don't affect enough people to be worth building UI or a workaround (Xcode also crashes if you move it and try to open or create a new project, for instance). They waste enough dev time to be worth detecting, but it's not urgent. One approach would be: - Watch the running versioned directory (with base::FilePathWatcher?). - If it moves or goes away, post a stat and set a crash key. - On the backend, ignore reports with that crash key. - Keep an eye on the stat.
,
Mar 7 2017
Users experienced this crash on the following builds: Mac Canary 59.0.3033.0 - 1.93 CPM, 3 reports, 3 clients (signature -[ToolbarController viewDidLoad]) If this update was incorrect, please add "Fracas-Wrong" label to prevent future updates. - Go/Fracas
,
Mar 8 2017
,
Apr 5 2017
Issue 703061 has been merged into this issue.
,
Apr 10 2017
,
Jun 22 2017
Issue 735454 has been merged into this issue.
,
Aug 27 2017
Issue 752047 has been merged into this issue.
,
Aug 29 2017
Just to update the latest behavior of this issue, still crash instances are observed on latest stable #60.0.3112.113. Below information provides the comparison between previous and latest channels including total number of instances. This crash is ranked as number #14 under Mac OS for browser process +--------------------------------------------------+ | Latest Channel | Previous Channel | |--------------------------------------------------+ |60.0.3112.113 - 79 | 60.0.3112.101 - 1016 |--> Stable +--------------------------------------------------+ Link to the list of the builds getting crash: --------------------------------------------- https://crash.corp.google.com/browse?q=product.name%3D%27Chrome_Mac%27%20%20AND%20custom_data.ChromeCrashProto.channel%3D%27%27%20AND%20custom_data.ChromeCrashProto.ptype%3D%27browser%27%20AND%20custom_data.ChromeCrashProto.magic_signature_1.name%3D%27-%5BToolbarController%20viewDidLoad%5D%27&ignore_case=false&enable_rewrite=true&omit_field_name=&omit_field_value=&omit_field_opt=%3D&unnest=#samplereports:5,productversion:1000 Thanks!
,
Sep 5 2017
Issue 761804 has been merged into this issue.
,
Oct 19 2017
Issue 775853 has been merged into this issue.
,
Jul 26
Just to update the latest behavior of this issue in the latest channels: Magic Signature - -[ToolbarController viewDidLoadImpl] This issue starting seen on M61-61.0.3163.79. But, crashes are still observed on latest Stable #68.0.3440.75. Still seeing 18 crashes from 18 clients so far on latest Stable - 68.0.3440.75 on Mac OS. This crash is ranked as number #4 in 'Browser' Stable crashes. 68.0.3440.68 0.01% 1 - Beta 68.0.3440.75 0.12% 19 - Stable So far no crash instances are observed on latest Canary and Dev. Link to the list of builds: ------------------------- https://crash.corp.google.com/browse?q=product_name%3D%27Chrome_Mac%27+AND+expanded_custom_data.ChromeCrashProto.ptype%3D%27browser%27+AND+expanded_custom_data.ChromeCrashProto.magic_signature_1.name%3D%27-%5BToolbarController+viewDidLoadImpl%5D%27#-productname:1000,productversion:100,-magicsignature:50,-magicsignature2:50,-stablesignature:50,-magicsignaturesorted:50 As shrike@ last visit is >30 days, lgrey@ could you please take a look into this issue based on the GIT Blame results for the file 'toolbar_controller.mm'. Also, cc'ing avi@ from the 'Owners' list for the file 'toolbar_controller.mm'. Thanks!
,
Jul 26
Just to add a note - Crashes are seen consistently on all stable builds.
,
Jul 26
We're already in the process of removing NIBs entirely, so given the priority, I'm goin g to WontFix this. |
||||||
►
Sign in to add a comment |
||||||
Comment 1 by sdy@chromium.org
, Mar 7 2017