From: Cherniavsky Beni <cben AT crosswinds DOT net>
Newsgroups: comp.os.msdos.djgpp
Subject: Re: multitasking/threading in DJGPP
Date: Thu, 22 Jun 2000 09:55:56 +0200
Organization: Technion
Lines: 62
Message-ID: <3951C68C.97B8437A@crosswinds.net>
References: <Pine DOT SUN DOT 3 DOT 91 DOT 1000619173815 DOT 27216T-100000 AT is> <394F7CEF DOT C3B04F13 AT crosswinds DOT net> <200006202002 DOT XAA00101 AT alpha DOT netvision DOT net DOT il>
NNTP-Posting-Host: 132.68.54.8
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
X-Mailer: Mozilla 4.06 [en] (WinNT; I)
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Reply-To: djgpp AT delorie DOT com

Eli Zaretskii wrote:
> 
> > From: Cherniavsky Beni <cben AT crosswinds DOT net>
> > Newsgroups: comp.os.msdos.djgpp
> > Date: Tue, 20 Jun 2000 16:17:19 +0200
> >
> > Isn't there a way to check at run-time with what size of stack the
> > program is configured?
> 
> Yes, there is: look at the value of _stklen.

My newsgroups access is from campus, while DJGPP is installed&used at home.
It's my fault that I was too impatient to go home and dig things up myself.
I just opened the FAQ online and complained about the (wrong) way I understood
it.  ;-(

> > The idea is that the library should allocate
> > more stack per thread in programs needing much stack space.
> 
> What library? the DJGPP standard library, libc.a?  I don't think it's
> library's business to decide how much stack to allocate.  It doesn't
> have enough information for that, not in the startup code anyway.

I meant the preemptive multithreading lib that batchex is writing - you
advised him to use the default stack size of DJGPP programs for the per-thread
stack allocated by his lib and I suggest it would be better to snarf the actual
size used by libc (which is anyway decided by the program author/user, not
libc).
 
> > It would be nicer if
> > the startup code updated _stklen with the actual value it used.
> 
> It does; please try it and tell if it doesn't work for you.  (The
> startup code in crt0.S clearly updates _stklen with the actual value.)

Same as first; I didn't test it nor looked at he source - my fault.  The only
conclusion is that maybe you should note in the FAQ (15.9) that _stklen is
updated if a larger value from the stub is used - at least I didn't understand
it
that way and possibly others won't.  BTW, now I also checked the (online :-)
libc
reference for _stklen and it's not clear from there too.
 
> > Anyway a better solution might be to allow the stack size for the
> > thread as a parameter when creating a thread.
> 
> That won't help: the stack of a DJGPP program is allocated in its
> entirety during startup, and its size cannot be changed later.  So the
> maximum stack size must be determined statically at compile time, or
> changed by STUBEDIT.

Same as second; I meant the stacks the multithreading lib allocates per thread.
They are allocated somehow or the disscussion of their size would be
meaningless.

PS: accept my appreciation for the effort you spend on answering posts :-)

-- 
Beni Cherniavsky <cben AT crosswinds DOT net>
                 (also scben AT t2 in technion)
What's the exterior of a domain?
When one takes a bath we shout "Man overboard!".