Re: Problems with Processes

From: Erick Gallesio <eg_at_kaolin.unice.fr>
Date: Fri, 21 Apr 1995 15:47:44 +0100

> Sean Slattery writes
>
> We're having trouble using run-process. The command seems to make my computer
> thrash about vigourously for a few seconds and then kill stk - I believe
> what's happenning is that stk is chewing up virtual memory at an amazing rate
> (I started at 21M free) and the OS is killing it before all the virtual memory
> is gone. Anyone else seen this behaviour?

Do you use the corrected version of process.c? (a corrected version is
available
on kaolin.unice.fr:/pub/Contrib).

> Two other questions:
>
> a) Are process supposed to be removed from (process-list)? I'm confused by the
> following behaviour (on a SunOS 4.1.3)
>
> [stuff deleted]
>
> My reading of the manual (page 38) says that process-wait returns #f if the
> process was already terminated when the call was made, but that process-list
> only returns the list of processes that are currently running. Am I misreading
> the manual, or is this a feature?

This should work with new release.

>
> b) STk seems to want to convert the result of a get command on a listbox to a
> list. This means that we run into trouble if our listbox entries have comma's
> or brackets in them (they get converted into (unquote ...) and sublists
> respectively). The man page does not state that <pathname> get <index> returns
> a list. I think it would make more sense for STk to keep this return value as
> a string to allow the user the maximum latitute in inserting stuff into the
> listbox that they later want to retrieve. Comments?

This is an old feature/bug that I want to correct since a long time.
The problem was that correcting it implies to modify the inspector, which I
have not written. I have done the modification but this yeild a weird bug on
trace variable which I have not tracked down yet (and that I can reproduce
only in the
inspector!!!). However, next release will have this correction in it

>
> One more thing - the slib common-lisp-time stuff doesn't seem to work - it
> falls back to using the counter. Please find appended my first stab at code to
> provide STk will a real universal time using the Unix date command - I hope
> the output of date is identical over machines, I've tested this on Sun running
> Mach and SunOS 4.1.6, DecStation running Mach, and a PowerPC running AIX
> 3.2.5. It's not good scheme code, but it works :)
>

I have added stuff for allowing the usage of C-types as first class objects in
STk. One application of this is to allow the definition of POSIX.1 strutures
for representating time. A complete implementation of POSIX.1 functions is now
available in the posix.c extension. It will be in the next release.



                -- Erick
Received on Fri Apr 21 1995 - 15:48:38 CEST