[bug 618] marked fixed, but dstress bug_optimize_294 showing some failures still

[Brad Roberts]

Thomas Kuehne wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Brad Roberts schrieb am 2007-01-19:
>> http://d.puremagic.com/issues/show_bug.cgi?id=618
>>
>> builds and runs for me: C, F, H, K
>>
>> fails to build for me:
>>    D:  bug_optimize_294_D.d(7): Error: incompatible types for 
>> ((_D7dstress3run1b18bug_optimize_294_D10__T3FooTaZ3Foo6__initZ) != 
>> (_D7dstress3run1b18bug_optimize_294_D10__T3FooTgZ3Foo6__initZ)): 
>> 'Foo!(char)' and 'Foo!(byte)'
>>
>> segv's at runtime:
>>    G:  Program received signal SIGSEGV, Segmentation fault.
>> 0x0804a09b in _Dmain () at bug_optimize_294_G.d:7
>> 7               if ((Foo!(char)).init == (Foo!(char)).init){
>>
>>
>> So.. the dstress results for C, F, H, K don't seem to match what you're 
>> seeing, Thomas.  I'm running on a 32bit linux box and testing with dmd 1.00.
> 
> I've fixed a few  copy'n paste errors in the test cases. The differences
> were moste likely caused because you tested
> 
> dmd source.d && ./source ; echo PASSED
> 
> while it should have been
> 
> dmd source.d && ./source && echo PASSED || echo FAILED
> 
> ("dmd -run" can't be used because it will return "0" regardless of
> "./source"'s return code)
> 
> Thomas
> 
> 
> -----BEGIN PGP SIGNATURE-----
> 
> iD8DBQFFsfQwLK5blCcjpWoRAnCvAJ957q8fbneCNfKSReoZXEcfot0yxgCgo/j7
> VDj8lktygyGTsunvp1Vkcb0=
> =KWMf
> -----END PGP SIGNATURE-----

I wasn't using the dstress framework, just copy/paste the code for the 
failing test and ran it manually.  I didn't actually even check the 
return code, just that it produced no output (close to the first case, 
but not even that elaborate :).  Sorry, I must have been awfully sleepy 
when doing that as it's obvious that's not sufficient.

What I'm actually thinking about here is how to check dstress vs 
bugzilla to make sure that bugs that are marked fixed are _actually_ 
fixed.  And then make sure that all the failing tests have a 
corresponding bug filed (many just point to a news group posting, which 
is easily and for most probably lost in the distant past).

Have you put any thought into how to do this verification step 
automatically?

Later,
Brad