SEGV in dwarf2read.c -- gdb-7.2

Jan Kratochvil jan.kratochvil@redhat.com
Thu Nov 3 16:55:00 GMT 2011


On Thu, 03 Nov 2011 17:43:20 +0100, Tom Tromey wrote:
> Does the CU cache serve a useful purpose these days?

You already need to pre-read all the CUs (=load_cu) before starting to parse
them (=process_queue).  So some multi-CU storage needs to be there.  It is
true the cache should be at least significantly larger
(now dwarf2_max_cache_age == 5 generations) to have any effect, if it can have
any effect.


> I wonder whether we could just remove it.

But the code does not get much simplified without the cache IMO (sure one
should benchmark it), there is still needed tracking of the multipled read in
CUs etc.


Regards,
Jan



More information about the Gdb-patches mailing list