[Sbcl-devel] Threads and 0.8.6

[Daniel B. <da...@te...>]

I don't think I'd realised quite how late in the month it is - well,
obviously I hadn't, because I'm just warming up to get involved in
interesting threading stuff.

In the next few days I plan to add 

1) support for futex-based queuing in Linuxen that have it

2) the thread/gc fixes that were left in a patch in contrib/ in
0.8.5

3) Andreas' patch for SIGRTMIN (which is better that than the one I
posted)

4) some changes to the interaction between background threads and the
debugger: remove the variable that wait-for-debugger checks to see if
it should just run the first restart instead (.i.e. so that background
threads always wait for the debugger).  In general we can't say what a
suitable error handling strategy is for unattended computation (the
first restart migh work, or might loop, or might blow the stack
recursively) but if the user has an appropriate plan for his app, he
can write it himself and hang it off the new *invoke-debugger-hook*

5) if things go well, redefine QUIT to do something less monumentally
stupid (but I'm not yet sure what)


If anyone would like to dissuade me from any/all of this, please feel
free.  I plan that all should be safely gathered in by the 20th


-dan

-- 

   http://web.metacircles.com/cirCLe_CD - Free Software Lisp/Linux distro