Hi Michael!

> The SoundBlaster drivers hook 
> EMM386's interrupt descriptor table, as well as its GDT (per previous 
> release remarks) and redirect all INT 3's (debug interrupt) to a general 
> protection fault with error code 1ah IF an application is running in 
> virtual 8086 mode.  As a result, any attempts to run an application under a 
> debugger with the SoundBlaster drivers loaded will fail with a 
> GPF.

You mean: classic style breakpoints become impossible. This means that
either the SBLive people are very jealous about somebody who might read
their code or they use breakpoints for special purposes. I seem to remember
that my SB AWE 32 "redirect MIDI to EMU8000 music" driver used the NMI for
something, maybe similar purposes in some way.

> Further, applications such as NEWDES, which intersperse their code 
> with normally benign INT 3's will fail with a GPF.  To work around this 
> problem,  the SB option redirects GPFs with error code 1ah to INT 3.

Nice. One thing which I remember: XTree file manager had INT 3 (1 byte
opcode) as replacement for INT 21, both to make the code smaller and make
debugging harder. So that falls into the category (why NEWDES?).

> Finally, I request that anyone who has a problem with any application 
> crashing while HIMEM or EMM386 is loaded, where the problem occurs only 
> when those drivers are loaded, let me know a link I can download the 
> application for testing, or send it to me directly.

Yes! Me... Tyler and me figured out that there is some incompatibility
between "LBAcache loaded into EMM386-generated UMBs" and "SCSI drive
detection". Symptom: Scanning through drives 0x80..0x87 takes much longer
than usual and in the end LBAcache will conclude that there are no
harddisks. Sometimes even CD-ROM spin up while scanning. The strange thing
is that I only use int 13.4100.55aa / 13.08 (ignoring returned ES:DI). The
code is in LBAcache setup.asm "findnextgeom" ... "alldrives". There are
no POINTERS to LBAcache load location handed over, maybe SCSI BIOS does
some DMA between the stack and the disk??? HOWEVER, Tyler thinks that this
even happens with PC DOS EMM386 (but not with UMBPCI), and I think that
e.g. VDS (thanks for re-enabling the possibility) cannot help me because I
do not explicitly point to UMB anyway...!?

So Lynx does not work... Any hope for volunteers to compile Links / ELinks
in DJGPP for DOS? I have done some preliminary chatting with the maintainer
so we can point to the places which need porting. To get started, one can
compile a non-networked version which uses ANSI. Later one can get the
networking code from SSHDOS added. Links is a pretty modern text mode browser.
Virtual PC problem - sure hard to test. GEOS Ensemble Lite - can you explain
in what way it is FreeDOS incompatible? Are you sure that this is not another
magical "2027 changed everything" (compare Novell NetWare) incompatibility?
And MSD ...: The programs which included the evil "pretend DR DOS is not
compatible" AARD test in Win3.1 were: MN, SETUP, HIMEM, SMARTDRV, MSD and WIN.
But it has been removed from HIMEM/MSD in Win3.1 final and hidden in WIN.
However, I think FreeDOS passes AARD, as otherwise SETUP would refuse to work.

http://fringe.davesource.com/Fringe/NonZen_Companies/Microsoft/Tactics/1993.09.01.Locks_Out_DrDOS.html

PS: Important *** INTERSVR / INTERLNK *** update: I was stupid...
DEVICE=c:\ms\interlnk.exe /noscan /low /noprinter /v /lpt1
INSTALL=c:\ms\intersvr.exe /x=c: /v /lpt1
^^^^^^^ not device...
...is the right syntax. And it DOES WORK in FreeDOS, at least it does
not crash at once in DOSEmu (sorry, I would need 2 PCs and a link cable
to do more tests). Note that INTERSVR does not run in the background: The
"Interlnk Server" 1.10 just shows full screen transfer status information,
telling me that it offers A:/D:/E:/F:/LPT1/LPT2 for remote access (which
are FAT12,FAT12,FAT16,FAT32 respectively, but I assume it would use only FAT1x
style int 21 calls if you would actually access F:).


PPS: Actually IBM EMM386 refused to work properly at all on that SCSI machine.

Eric


-------------------------------------------------------
This SF.Net email is sponsored by BEA Weblogic Workshop
FREE Java Enterprise J2EE developer tools!
Get your free copy of BEA WebLogic Workshop 8.1 today.
http://ads.osdn.com/?ad_id=4721&alloc_id=10040&op=click
_______________________________________________
Freedos-devel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/freedos-devel

Reply via email to