Re: Another unwinding problem on ARM

[Mathias Bauer]

Hi,

as I found out, the EHT entries generated for Objective-C code claim 
that the TTypeEncoding is DW_EH_PE_absptr, but in fact the encoding of 
the pointers in the type table is DW_EH_PE_pcrel | DW_EH_PE_indirect.

I don't know if this is an ARM-specific treatment that the personality 
routine must be able to cope with or if this has to be fixed in the 
generated EHT. I leave this up to the experts. In case it will be 
decided that we have to fix that in the runtime, I attached a possible 
patch.

Best regards,
Mathias

Am 14.03.14 16:58, schrieb Mathias Bauer:

> > Now the test case works - at least when run inside gdb. I still have to
> > find out why it core dumps without gdb.
>
> And this turns out to be a problem of interpreting the LSDA.
>
> The stack frame in question belongs to [NSOperation start].
> When the PR __gnu_objc_personality_v0 is called for it, the
> internal_objc_personality function calls check_action. The
> resolve_indirect_value function returns something that - when
> interpreted as a char* - leads to a crash.
>
> I wonder whether the internal_objc_personality needs some adjustments
> for _US_FORCE_UNWIND here.
>
> I will continue investigating this, but I would appreciate any hints
> that could help me speed this up. My wife probably will get angry if it
> takes the whole weekend. ;-)
>
> Regards,
> Mathias
>
> _______________________________________________
> Discuss-gnustep mailing list
> Discuss-gnustep@gnu.org
> https://lists.gnu.org/mailman/listinfo/discuss-gnustep

dwarf_eh.diff
Description: Text document