This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: rs6000 SPE breakage

Hi,
sorry for not responding till now, but I already left to the trip.
> 	If I remove the call to htab_expand() in htab_traverse(), the ICE
> does not occur.  Why is it safe to compact the hash table at the beginning
> of a call to htab_traverse?  It looks like cselib may save some
> information in locs that it expects to remain consistent between its two
> calls to htab_traverse() which would break if the compaction occurs at the
> beginning of the second call.

Cselib is definitly not doing that.  It is not safe to save pointer as
the table is rehashed at insertion...
It should not be that dificult to find it.  Only way to get pointer
into hash table bucked is htab_find_slot.
Is the crash inside clelib?

Honza
> 
> David
Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]