Re: [Gc] Re: Re[4]: [bdwgc] Exit abort hooks

Ivan Maidanski ivmai at mail.ru
Thu Jul 5 01:14:06 PDT 2012


Hi Jean-Claude,

Please test the code in this branch: https://github.com/ivmai/bdwgc/tree/add-signals-setters
It should have the functionality you requested, here is the diff (it contains some code fragments from your patches): https://github.com/ivmai/bdwgc/compare/add-signals-setters

Regards,
Ivan

Tue, 3 Jul 2012 02:18:03 -0400 Jean-Claude Beaudoin <jean.claude.beaudoin at gmail.com>:
> On Sat, Jun 30, 2012 at 11:50 AM, Ivan Maidanski <ivmai at mail.ru> wrote:
> 
> > Rejected. Looks to me incorrect - it seems you just add some variables and
> > setters for them but the variables are not used in GC (e.g., resume signal
> > is always SIG_THR_RESTART).
> >
> 
> If you have a look in my github repository (jcbeaudoin/bdwgc, branch
> signals_init) at
> file "include/private/gc_priv.h" around lines 2290-2293, you will see a
> macro definition like this one:
> 
> #     define SIG_THR_RESTART GC_get_thr_restart_signal()
> 
> This macro redirection will replace every occurrence  SIG_THR_RESTART by a
> call
> to the new GC_get_thr_restart_signal() accessor thus providing the
> init-time configured
> signal value.
> 
> BTW, in the pull request I last sent you, there was a typo in the name of
> the new restart
> signal setter function; it should have read GC_set_thr_restart_signal()
> instead of
> GC_set_thread_restart_signal().
> 
> Short if this typo the code I submitted to you runs fine on Linux as is,
> but did not extend the
> new functionality to all other signal dependent platforms. For illustrative
> purposes I did the
> necessary modifications to implement this extension in a later commit to my
> repository.
> It mainly concerns file "include/private/gcconfig.h" and took only a few
> minutes to do.
> Since I have no access to those platforms outside Linux on Intel x86 I have
> not been
> able to compile/run this last modification on any of them.
> 
> 
> > Let's start from the beginning: you want some new GC functionality
> > allowing to suggest the collector to use some specific signals for POSIX
> > threads suspension and/or resuming. To make this, I expect from you several
> > easy-to-review commits:
> >
> 
> Your demands far exceed the amount of efforts I can reasonably devote to
> this exercise.
> 
> As far as I am concerned this is the end of this conversation.
> 
> So long, and thanks for all the fish. :-)




More information about the Gc mailing list