This is the mail archive of the gdb-patches@sourceware.org mailing list for the GDB 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: [patch] STT_GNU_IFUNC support


On Wed, 17 Feb 2010 18:55:10 +0100, Pedro Alves wrote:
> On Wednesday 17 February 2010 17:33:38, Jan Kratochvil wrote:
> > > > What about making this GNU-IFUNC inferior call scheduling follow the "step"
> > > > policy?  Maybe the whole inferior calls should follow the "step" policy?
> > > 
> > > Or "on", should be the same.
> > 
> > Not so.  I find "step" to be the reasonable default (and it has been so for
> > a long time before me in RHEL/Fedora) and I find GNU-IFUNC resolving with
> > locked scheduler also as a reasonable default.  I do not find
> > "scheduler-locking on" as a reasonable GDB default.
> 
> Oh, of course not.  I was answering the first question,
> about the specific infcall to resolve "strcmp" when the
> user did "p strcmp" or "b strcmp".  I assume you meant to
> sched-lock that call, which would be the same as "on",
> but maybe I misunderstood what you meant.

We agree on the significant part.  The other part is offtopic for IFUNC.


Thanks,
Jan

------------------------------------------------------------------------------

> A reply to
> the second question would be similar to
> the "That's another story." reply.  ;-)

The "That's another story." part:

To make it whole more clear I would suggesting creating:

set scheduler-locking-continue (on|off)
set scheduler-locking-step (on|off)

mapping current settings -> the new scheme as:

set scheduler-locking on   -> s-l-continue=on  s-l-step=on
set scheduler-locking step -> s-l-continue=off s-l-step=on
set scheduler-locking off  -> s-l-continue=off s-l-step=off

and introducing for infcall/gnu-ifunc new settings:

set scheduler-locking-infcall (on|off)
set scheduler-locking-gnu-ifunc (on|off)

Currently s-l-infcall behaves the same as s-l-continue.
IMO s-l-infcall should behave more the same as s-l-step.


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