Re: Tcl and rthreads.
On 03/24/12 16:53, Stuart Cassoff wrote: > I'd like to (eventually) make threaded Tcl the default > and have a nonthreaded flavor. > > Currently experiencing the following test failures > which didn't happen before rthreads: > > async-4.3 fails on i386 and amd64 (sporadic) > event-14.1 fails on i386 (sporadic) > interp-36.7 fails on i386 and amd64 > > Also, unixFCmd-15.2 fails on amd64. > This is a Tcl problem that has been fixed in the Tcl HEAD. > I'll cook up a patch for that soon enough. > > Also also, test failures fCmd-9.4.b and stack-3.1 > are to be expected. This is ok. > > The problem seems to be something with timeouts. > Maybe select() related? > > Not sure where to go from here and of course I'll be > happy to assist with any debugging, testing patches, etc. > > Stu > Situation update: All thread-related issues mentioned have been fixed except for whatever is causing the interp-36.7 failures. Here are the gory details: http://sourceforge.net/tracker/?func=detail&aid=3544685&group_id=10894&atid=110894 Any help at all on this would be greatly appreciated since it seems to be the last problem holding back a move to threaded Tcl. It's all a bit beyond me and there aren't many other tcl/obsd people in tcl-land who have the time/skill to look into this, so, yeah, as mentioned, help would be great. I'm available for testing and trying out stuff. Thanks, Stu
Re: Tcl and rthreads.
2012/3/25 Jeremy Evans : > On Sat, Mar 24, 2012 at 1:53 PM, Stuart Cassoff wrote: >> The problem seems to be something with timeouts. >> Maybe select() related? > > I've also seen a ruby 1.9 regress failure related to select after the > upgrade to rthreads. I haven't had time to look into it yet, though. While on the topic of threading ports, I made a patch for the erlang port: When starting "erl" with this port, it prints -> [smp:2:2] [rq:2] on a dual-core as expected Only tested on amd64. Hope gmail won't make the diff all worthless. It's easy enough to retype I think. I have verified that a 2-core laptop does fibonaccis twice as fast with this patch, have no idea how to test it more than this (ie. I don't have a decent ejabberd or other env. to haunt, the port doesn't have a regress target) cvs diff: Diffing . Index: Makefile === RCS file: /home/cvsync/cvs/ports/lang/erlang/Makefile,v retrieving revision 1.46 diff -u -r1.46 Makefile --- Makefile23 Dec 2011 13:40:13 - 1.46 +++ Makefile26 Mar 2012 07:55:14 - @@ -6,6 +6,7 @@ DISTNAME= otp_src_${V} PKGNAME= erlang-14b.04 EPOCH= 0 +REVISION= 0 CATEGORIES=lang # Erlang Public License @@ -32,7 +33,7 @@ --enable-threads\ --enable-kernel-poll \ --disable-hipe \ - --disable-smp-support + --enable-smp-support USE_GMAKE= Yes USE_GROFF =Yes cvs diff: Diffing patches cvs diff: Diffing pkg Index: pkg/PLIST === RCS file: /home/cvsync/cvs/ports/lang/erlang/pkg/PLIST,v retrieving revision 1.21 diff -u -r1.21 PLIST --- pkg/PLIST 23 Dec 2011 13:40:13 - 1.21 +++ pkg/PLIST 22 Mar 2012 19:49:06 - @@ -191,6 +191,7 @@ lib/erlang/${ERTS}/ lib/erlang/${ERTS}/bin/ @bin lib/erlang/${ERTS}/bin/beam +@bin lib/erlang/${ERTS}/bin/beam.smp @bin lib/erlang/${ERTS}/bin/child_setup @bin lib/erlang/${ERTS}/bin/ct_run @bin lib/erlang/${ERTS}/bin/dialyzer -- To our sweethearts and wives. May they never meet. -- 19th century toast
Re: Tcl and rthreads.
On Sat, Mar 24, 2012 at 1:53 PM, Stuart Cassoff wrote: > The problem seems to be something with timeouts. > Maybe select() related? I've also seen a ruby 1.9 regress failure related to select after the upgrade to rthreads. I haven't had time to look into it yet, though. Jeremy
Re: Tcl and rthreads.
On Sat, Mar 24, 2012 at 9:53 PM, Stuart Cassoff wrote: > I'd like to (eventually) make threaded Tcl the default > and have a nonthreaded flavor. Threaded perl would also be nice... ciao, David
Re: Tcl and rthreads.
On 03/24/12 17:23, Brad Smith wrote: > On 24/03/12 4:53 PM, Stuart Cassoff wrote: >> I'd like to (eventually) make threaded Tcl the default >> and have a nonthreaded flavor. > > The whole point of me getting you to enable this by default was to > remove the FLAVOR use all together not to add another one. > No problemo. After I posted I was thinking that the nonthreaded flavor wasn't necessary. Still doesn't fix the problems with threads, though. Stu
Re: Tcl and rthreads.
On 24/03/12 4:53 PM, Stuart Cassoff wrote: I'd like to (eventually) make threaded Tcl the default and have a nonthreaded flavor. The whole point of me getting you to enable this by default was to remove the FLAVOR use all together not to add another one. -- This message has been scanned for viruses and dangerous content by MailScanner, and is believed to be clean.