Re: [MiNT] kernel 1.15.10b fragmentation

[Guido.Flohr@t-online.de (Guido Flohr)]

Hi,

On Tue, Jan 02, 2001 at 03:29:24PM +0100, Petr Stehlik wrote:
> On Tue, 2 Jan 2001, Thomas Binder wrote:
> 
> > > replacements yet and their source code is not available (ST-Guide has
> > > been recently named as one of those problematic but widely used
> > > programs). The proposed kernel call doesn't help them, does it?
> > 
> > Nope. But it helps to set the memory protection flags of such offending
> > programs to `readable' in most cases - sometimes you need `global',
> > depending on what protocol messages they use.
> 
> Sounds like the proposed kernel call would allow somebody to write a
> checker - an application that would automatically scan your system, find
> all offending programs and set their memory protection flags. At least
> this is something that would be really helpful. Is it possible to automate
> this task? I doubt that.

That could already be done with the existing solution
(SIGBUS-handler).  But my intention was actually to protect mp-friendly
(new!) programs from the black sheep.

> Another approach is to manually keep a list of all programs together with
> suggested memory flags. Then a simple script could be written that would
> again scan your system, find the files and set their flags. Just to have
> the list of all programs :-)

I think a black list of programs that need to have their program flags set
to global would be sufficient.  In fact that list would not be too large,
on my MiNT system only a handful of applications needed to be
reflagged.  But kind of an official list with problematic programs
(including version number if the problems were fixed during
development) would surely be handy.

Another quite unconventional proposal: What about some facility that
allows "clean" programs to report and log failed calls to Maccess?  It
would probably be helpful if users could obtain a list of problematic
programs from some logfile like /var/log/mpswines. ;-)

Even less conventional: The MiNTLib could be configured to do that
automatically.  This is a little time-consuming but this feature would
probably only be used for debugging purposes.

Ciao

Guido