hans.boehm at hp.com
Wed Jun 6 13:11:17 PDT 2007
I finally checked in your last patch. I needed to make one correction
since the flags field and FINISHED macro are not defined in a pure
I also finally switched to a machine with newer Windows tool chains,
including the Microsoft 2005 win64 tool chain. I already checked in a
bunch of mostly minor patches as a result of trying to get those to
work. I will probably be doing more of the same.
> -----Original Message-----
> From: Romano Paolo Tenca [mailto:rotenca at telvia.it]
> Sent: Saturday, May 26, 2007 11:39 AM
> To: Boehm, Hans
> Cc: Gc at napali.hpl.hp.com
> Subject: Re: [Gc] win32_pthreads
> Boehm, Hans ha scritto:
> > I haven't had a chance to look at the patch yet, but this makes me
> > suspicious that there is a mingw/win32_pthread internal
> lock that we
> > should be acquiring before stopping the world, at least if we want
> > GC_printf to work with the world stopped. Of course, that
> may well be
> > impossible, for example, because it may not be visible to
> clients ...
> The thread native version uses the CRITICAL_SECTION
> GC_write_cs to arbitrate stop_world and GC_write, I think
> that the same should be done for pthreads.
> But if the user is suspended inside a WriteFile, no
> arbitration will be done, so the stack debug print should be
> used with attention and decoupled from standard debug prints.
> The last addition of the declaration of GC_beginthreadex in
> gc.h require the inclusion of <stdint.h> under mingw.
> Romano Paolo Tenca
More information about the Gc