Re: Topics about GNUstep/libobjc2

[Dr. Rolf Jansen]

Hi David!

> Am 25.10.2016 um 12:28 schrieb David Chisnall <theraven@sucs.org>:
> 
> It would help if the reduced test case compiled, but adding -march=native 
> seems to fix that.

On my FreeBSD machine, -march=native doesn't help.

   Intel(R) Atom(TM) CPU D510  @ 1.66GHz (1666.72-MHz K8-class CPU)

> It would also help if it were something a bit more minimal, but from what 
> you’ve sent me:

Please excuse this. I started creating a test case from scratch, but that one 
didn't care to crash. Then I thought that the "ambient" of the incriminated 
piece of code does matter, and so I stripped down the actual project and let 
the involved structs and types in place, because their offsets might have made 
up for the difference. After this exercise, I ran out of time and for this 
reason I could not reduce the test case further -- I was just happy that I 
could present something that shows the problem.

> The crash happens in [CalcFactor evaluate], which simply returns the 
> ResultType ivar itsValue.

Yes, accessing itsValue was consistently the problem in the actual project.

> Clang is emitting sse instructions for this, assuming that the ivar is 
> correctly 128-bit aligned.  The offset is 48, so it is aligned within the 
> object, but the 64-bit header makes it unaligned again.

Well, I am glad that you found the problem, and tonight I will try out libobjc2 
from GitHub.

Again, thank you very much for taking your time.

Best regards

Rolf