Message: previous - next
Month: May 2012

Re: [trinity-devel] kwrite Crash - Now Down To 3/29 -> 4/1 (this may not be good -- gcc to blame?)

From: "Timothy Pearson" <kb9vqf@...>
Date: Wed, 2 May 2012 20:15:06 -0500
>> Tim, Darrell, All:
>>   I rebuilt TDE from 4/1 sources with full debug in attempt to further
>> look into
>> the kwrite crash that Darrell and I have experienced with TDE built on
>> gcc
>> 4.7.
>> With this build, if it is a source/bad patch/tde issue, then with my
>> 3/29
>> build
>> giving no crash and my 4/1 build crashing, that would point to some
>> problem in
>> between those two dates. However, I am coming around to the thought that
>> -- this
>> isn't a TDE problem and is more likely a problem introduced by the gcc
>> 4.7
>> interpretation of the TDE code. I will still have to build the 3/29
>> sources with
>> gcc 4.7 to confirm that to be the case. I have 3/29 binaries that work
>> fine
>> (which I'm fairly sure were built on gcc 4.7, but I can't remember). If
>> I
>> rebuild with gcc47 on the 3/29 sources and get the same failure -- that,
>> in my
>> mind, conclusively points to the problem being gcc47.
>>   If that is the case -- how in the heck do you debug something like
>> that?
> Upgrading gcc should NOT cause working code to start crashing unless the
> C/C++ standard was violated somewhere.  As ugly as it may be, we may need
> to entertain the notion that gcc 4.7 is actually generating buggy code (it
> would not be the first time, IIRC gcc 4.3 caused crashes as well that were
> magically resolved with gcc 4.4).
> That said, I hope you find something soon and that this is NOT a gcc
> related issue!
> Tim

That should read that gcc 4.4, not 4.3, was a problem.