Message-Id:	<m0w0Tn7-000S1nC@natacha.inti.edu.ar>
Comments:	Authenticated sender is <salvador AT natacha DOT inti DOT edu DOT ar>
From:	"Salvador Eduardo Tropea (SET)" <salvador AT natacha DOT inti DOT edu DOT ar>
Organization:	INTI
To:	pekangas AT sci DOT fi (Petteri Kangaslampi), djgpp AT delorie DOT com
Date:	Fri, 28 Feb 1997 18:13:00 +0000
MIME-Version:	1.0
Subject:	Re: MIDAS 0.6 for DJGPP beta has been released

Hi All about the MIDAS topic:

> On 28 Feb 1997 11:37:30 GMT, Esa A E Peuha <peuha AT cc DOT helsinki DOT fi> wrote:
> >Petteri Kangaslampi (pekangas AT sci DOT fi) wrote:
> >: we don't lock any of the memory we use, even though most of it gets used
> >: in the timer interrupt. Everything seems to work OK, both under DOS
> >: (cwsdpi) and Win95, but this still isn't good practise. Any ideas on how
> >: to easily lock all memory used?
> >
> >A timer interrupt handler shouldn't be too difficult to write in
> >assembly (which will be easy to lock) especially since you already have
> >the equivalent C code. Just put the handler (and any other code it could
> >[...]
> 
> Yes, but the problem is that practically all code in MIDAS can get called
> inside the interrupt, as all music playing is done there. That's why I
> have been looking for a good way to determine the whole memory area used
> by the library, and lock it as a one big chunk. So far the best idea I
> have had is to add two dummy modules with labels in all segments (code,
> data, BSS), add them as first and last modules, and use them to determine
> the memory to lock. The problem is, this depends on how the linker orders
> the modules, and doesn't sound too reliable because of that.
> 
> Meanwhile, MIDAS seems to work very reliably without locking, I haven't
> had a single crash related to that myself, and neither have I heard of
> anyone else. The problems probably start when cwsdpmi starts swapping to
> disk.
> 
> BTW, everybody, if the timer borders get too annoying, mail me for
> instructions on how to remove them :) The next MIDAS release will probably
> be the MIDAS Juhla Version, out Friday the 7th, with quite a few updates
> and fixes. DOS only.
> 
> 
> Petteri

As I told to Petteri before there are an automatic way to group the code and 
data that must be locked, lamentably the patchs for gcc to achieve that needs 
to be aproved(?) by the FSF and until now Bill Curie and me didn't get any 
replay from the FSF.

SET  
--------------- 0 --------------------------------
Salvador Eduardo Tropea (SET).
Address: Curapaligue 2124, Caseros, 3 de Febrero
Buenos Aires, (1678), ARGENTINA
TE: +(541) 759 0013

- Raw text -