Open
      
        Bug 1241166
      
      
        Opened 9 years ago
          Updated 3 years ago
      
        
    
  
DMD eventually crashes when running sequential torture mode in emunittest suite. 
    Categories
(Core :: DMD, defect)
Tracking
()
        NEW
        
        
    
  
People
(Reporter: jujjyl, Unassigned)
Details
Attachments
(3 files)
STR:
1. Build Firefox with DMD enabled.
2. Run the emunittest 0.5 suite in "Run all tests" mode with "sequential torture mode" enabled.
After a while, Firefox content process will crash in dmd.dll, e.g. https://crash-stats.mozilla.com/report/index/639d2a2b-fa73-4e78-b907-53c992160120
Please contact me directly to obtain the emunittest suite.
|   | ||
| Comment 1•9 years ago
           | ||
The crash has no source file/line info. Is it possible to get a crash that does have that info?
Flags: needinfo?(jujjyl)
| Reporter | ||
| Comment 2•9 years ago
           | ||
I'm not sure why it doesn't have that info and what one has to do to add it. I'll see if I can get the crash again to see if it would report anything better.
Flags: needinfo?(jujjyl)
| Comment 3•9 years ago
           | ||
It isn't including that info because you probably used a local build, so there are no symbols for it on crash-stats. You could try running the tests under a debugger. That might give you a better stack trace when it crashes.
| Reporter | ||
| Comment 4•9 years ago
           | ||
Disabled e10s and started Firefox with MOZ_CRASHREPORTER_DISABLE=1, and ran it attached to Visual Studio. Running emunittest suite repeatedly eventually crashed in DMD as shown in this callstack.
| Reporter | ||
| Comment 5•9 years ago
           | ||
Stack of the parent function looks like this.
| Reporter | ||
| Comment 6•9 years ago
           | ||
And the parent of that function.
Assuming that Visual Studio debugger interpretation is correct, then gLiveBlockTable->lookup() returned an object p, for which p->Report() has a bogus this pointer, which would indicate that the LiveBlockTable::Ptr p that was returned was bogus. Not sure for certain though.
| Reporter | ||
| Comment 7•9 years ago
           | ||
Interestingly the crash occurred when loading Dead Trigger 2 in the suite, which is available also at http://beta.unity3d.com/jonas/DT2/, and the first time the crash occurred, it was also during loading Dead Trigger 2, and on both runs, it occurred on the second run-through of the suite, so first Dead Trigger 2 load succeeded ok.
| Updated•3 years ago
           | 
Severity: normal → S3
          You need to log in
          before you can comment on or make changes to this bug.
        
 crash_callstack.png
 crash_callstack.png
            
Description
•