[Openmcl-devel] Unix signal handling

[Daniel Weinreb]

You added  ccl:*quit-interrupt-hook*, which is called
on SIGTERM.  Does it have these problems? It might
not be that important for us, since what we're going
to do is a shutdown anyway, but, if it's going to be
flaky, I'd at least like to know what we're up against.
Thanks.

-- Dan

Gary Byers wrote:
>
>
> On Mon, 5 Jul 2010, Scott L. Burson wrote:
>
>> Hi,
>>
>> I am porting a large, old application from Allegro to Clozure.  The 
>> app runs
>> under Emacs -- it has its own code, on both the Emacs and Lisp sides, 
>> for
>> doing this -- and Emacs sometimes gets the Lisp process's attention by
>> sending it a Unix signal (normally SIGUSR1 or SIGUSR2, but this is
>> configurable).
>>
>> Allegro provides a relatively straightforward way to set up a Lisp 
>> function
>> to be called when a signal is received, but I don't see anything like 
>> that
>> in CCL.  Have I just overlooked it?  If not, how hard would it be to 
>> create?
>
> There isn't any general way of defining signal handlers in CCL.  The 
> issue
> comes up every few months on this list, and the response is usually 
> "yeah,
> we should offer that.  Someday."  It's almost certainly doable, but 
> there are
> threads/GC issues that make the implementation complicated.
>
> Additionally, the implementation of things like PROCESS-INTERRUPT and the
> mechanism that the GC uses to suspend/resume other threads involves 
> sending
> signals between threads; on some platforms (Darwin), the 
> implementation uses
> SIGUSR1 and SIGUSR2 and possibly something else that the OS never 
> raises, and
> it's not clear that there are any leftover unused signal numbers ...
>
> It might be tempting to just do:
>
> (defcallback my-signal-handler (:int signo :void)
>   (some-code-that-can-run-at-interrupt-time))
>
> (#_signal #$some-signal my-signal-handler)
>
> and that'd work some of the time.  It'd fail (subtly or spectacularly) 
> if the
> GC ran in some other thread and that signal handler interrupted some 
> lisp code.
> (The GC would likely not be able to see the lisp object references 
> that that
> interrupted code was making, and things would go downhill from 
> there.)  There
> are some other issues there, so any viable mechanism would have to 
> deal with
> those issues around a call to your code.
>
>
>>
>> Alternatives exist, of course -- I could set up another stream with a 
>> Lisp
>> thread reading from it, and just send a character on the stream.  Sounds
>> like a bit of work, though I haven't looked into it; not sure if 
>> Emacs will
>> cooperate.  Anyway I wanted to explore the signal handling approach 
>> first,
>> since that is how it works with Allegro.
>>
>> Any advice?
>>
>> -- Scott
>>
>>
>>
> ------------------------------------------------------------------------
>
> _______________________________________________
> Openmcl-devel mailing list
> Openmcl-devel at clozure.com
> http://clozure.com/mailman/listinfo/openmcl-devel
>   
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.clozure.com/pipermail/openmcl-devel/attachments/20100706/9cf6d793/attachment.htm>