Path: utzoo!censor!geac!torsqnt!news-server.csri.toronto.edu!cs.utexas.edu!sun-barr!lll-winken!elroy.jpl.nasa.gov!sdd.hp.com!spool2.mu.edu!uunet!cbmvax!cbmehq!babylon!rbabel
From: rbabel@babylon.rmt.sub.org (Ralph Babel)
Newsgroups: comp.sys.amiga.programmer
Subject: Re: Bugs in 2.04
Message-ID: <05773.AA05773@babylon.rmt.sub.org>
Date: 22 Jan 91 04:58:22 GMT
References: <91022.110834GHGAQA4@cc1.kuleuven.ac.be>
Reply-To: cbmvax.commodore.com!cbmehq!babylon!rbabel (Ralph Babel)
Lines: 36

In article <91022.110834GHGAQA4@cc1.kuleuven.ac.be>
GHGAQA4@cc1.kuleuven.ac.be (Tyberghein Jorrit) writes:

> I have AmigaDOS 2.04 and I discovered some alarming bugs :

We'll see about that ...

> The list of locks for the RAM disk is corrupt (my debugger
> and Xoper crash when they try to list them). This was not
> the case in AmigaDOS 2.02.

The content of fl_Link in case of volumes that are currently
mounted is undefined. Since the RAM disk doesn't support
removable media (otherwise it would have to generate a
DISKREMOVED event when a RAM expansion is taken out of the
machine :-), there can't be a RAM disk volume with
outstanding (i.e. linked) locks.

> Some processes (like DF0, RAM, WB_2.x, HARD) all have a
> corrupt process seglist. Again my debugger and Xoper give
> wrong information when trying to look at these. This was
> also not the case in AmigaDOS 2.02. All these process have
> something to do with disk devices. Could this be the
> problem ?

Even though the contents of pr_SegList indeed appear to be a
bit messed up (Randell?), again the real problem is in
relying on undocumented side-effects. In case of GlobVec==0,
DeviceProc()/LoadDevice() require pr_SegList (really:
SegArray) to inherit global vector addresses both from the
DOS kernel and the application (good old BCPL loader-format
- those who have written handlers before 1.2 know what I'm
talking about :-). BCPL is (almost) history, so don't count
on all that undocumented magic.

Ralph