Re[2]: [Gc]: Request: get signum for resume?

Ivan Maidanski ivmai at mail.ru
Thu Dec 15 08:15:32 PST 2011


Hi,

1. name: GC_get_thr_restart_signal // in misc.c; same behavior as for suspend.

2. SIG_THR_RESTART: move to gc_priv.h; try to not define it unless used in pthread_stop_world.c

3. Send me initial patch - if something is wrong, I'll either fix & commit or send you feedback.

Regards.

15 12 2011, 19:51 Jan Wielemaker <J.Wielemaker at vu.nl>:
> Hi Ivan,
> 
> On 12/15/2011 04:24 PM, Ivan Maidanski wrote:
> > Hi Jan,
> >
> > Sure. Please prepare a patch (against master branch in Git repo).
> 
> Fine.  Only, it is a bit of a mess :-(  I guess I must add this to
> misc.c, but the definition of SIG_THR_RESTART is spread over
> gcconfig.h and pthread_stop_world.c :-(  Shall I move the part
> from pthread_stop_world.c to gcconfig.h and add the new function
> on misc.c?  What name to use?  resume, restart or thr_restart?
> 
> More, I see that pthread_stop_world.c is only used under some
> complex condition.  Should we move that condition to gcconfig.h
> as well and define a new GC_PTHREAD_STOP_WORLD that is used as
> a condition in pthread_stop_world.c?
> 
> I'm happy to write a bit, but it not my code base ...
> 
> 	Cheers --- Jan
> 
> > Regards.
> >
> > 15 12 2011, 17:37 Jan Wielemaker<J.Wielemaker at vu.nl>:
> >> Hi,
> >>
> >> Just making SWI-Prolog run under bwdgc wasn't too hard! I ran against
> >> one glitch that can easily be fixed. It turns out the application should
> >> not play with the collector's signals. Well, that is documented and not
> >> a surprise. Prolog however catches most signals to generate a Prolog
> >> exception.
> >>
> >> I found GC_get_suspend_signal(), which gives me a nice portable way to
> >> leave that one unmodified, but I must do the same with SIGXCPU (test
> >> platform is Linux) and reading the source indicates there is no
> >>
> >> 	GC_get_resume_signal()
> >>
> >> Could this be added?
> >>
> >> 	Thanks --- Jan
> >>
> >> P.s.	Time for extensive testing and evaluation of memory usage and
> >> 	performance.  I'm very curious.  If it works out I can ditch
> >> 	a lot of code :-)
> 
> _______________________________________________
> Gc mailing list
> Gc at linux.hpl.hp.com
> http://www.hpl.hp.com/hosted/linux/mail-archives/gc/
> 



More information about the Gc mailing list