Does Amix issue a REQUEST SENSE after the first REWIND? If so, are you returning the BOM (beginning of medium) bit set in the request sense data then?

I don't see any BOM-bits in sense data, only end of media (EOM) bit.

But yes, sequence is:
REQUEST SENSE
REWIND
REQUEST SENSE
REWIND
REQUEST SENSE

I still don't have any ideas.. But here is test version if anyone cares. http://www.winuae.net/files/b/winuae.zip

Tape emulation works currently strangely, all files must be in single archive (for example zip) and file called tape.index must be also included that contains order of files.

Manually add something like following to config file to add tape drive (uaehf0 = SCSI HD):

uaehf1=tape0,rw,ABC:<path to tape.zip>,0,0,0,512,0,,scsi4
(ABC: is unused device name. Apparently tape drive must use SCSI ID 4)

tape.index contents when using 2.01 install:

What if you set the EOM bit when a REWIND is attempted and the tape is already positioned at the start? The SCSI (1) standard doesn't say whether that should definitely happen or not though:
There are a couple of tape drive manuals which might be of use:
Wangtek 5525ES (I think the earlier 5150ES might have worked with Amix, maybe that drive is backward-compatible?)
Archive Viper 2060S/2150S

I remember reading of a patch to Amix to get it to work with the Viper drive. So you probably couldn't use a Viper to install Amix using unmodified boot disks.

It seems the QIC-150 format only specifies 512-byte blocks. Nothing else is allowed.

In the Wangtek 5525ES PDF, see the REQUEST SENSE details starting at page 82. That mentions a BOM bit (byte 9 bit 3 in the request sense data), and also ASC/ASCQ of 00/04 meaning "Beginning of Medium or beginning of partition detected". The drive can also return SCSI-2 extended sense data, where BOM is byte 19 bit 3. Whether the drive is in SCSI-1 or SCSI-2 mode is determined by a jumper on the PCB.

Also note the amigasrc and pubsrc archives should be compressed with the UNIX compress command (as they are on the original tape). So amigasrc.cpio.Z and pubsrc.cpio.Z would be better names for those.

Setting BOM bit seems to help, now it does not abort immediately anymore but attempts to execute MODE_SELECT command (which is not yet implemented).

I guess these bits are obsolete because "SCSI Stream Commands" document from 2003 does not mention them. (Wangtek manual looks nice, thanks)

Amix only reads 14 bytes of sense data, I guess it means Amix only works with SCSI-1 tape drive.

Tape installation works! Winuae.zip updated.

BOM-bit and some fixes in SCSI emulation needed. Wrong data size if command data direction was from Amiga to device and command was not write (MODE SELECT) and WD Select and Transfer was used. Yet another situation that was not used by any other SCSI driver.

EDIT: I only tested 2.01 tape set.

Tape drive emulation is now mostly finished, including GUI support.

Added plain directory support, archives are still supported. Index file renamed to index.tape.

Ho-ly crap.

AGA support, A3000 chipset support, A2024 support, MMU support, and now tape. My life is complete...and so is the A3000UX.

*bows gracefully at Toni*

Rodney

Hardware level graphics card emulation is still missing :)

I was listing out as complete a list I could remember of things that Toni said he'd never do...and then did anyway, much to our benefit. =)

Rodney

Are you referring to the A3000UX? Did it come with a dedicated card? (I always thought it was "software" out to the A2024 - but have never seen one in real life.)

Rodney

I am quite sure I never said I'd not emulate AGA because when I started with UAE, AGA was one of the first things I partially implemented :)

Amix has drivers for some "real" graphics cards. IMHO A2024 is not a graphics card, more like a hack/workaround for chipset limits. (Graphics board emulation plans: http://eab.abime.net/request-uae-wis...emulation.html)

I've been playing with AMIX for the last dozen days, thanks so much Toni for getting the MMU emulation working! I'm posting here: http://forums.nekochan.net/viewtopic.php?f=8&t=16727870. I'm trying to add a few more tools to update some of the most glaringly outdated programs. Especially for DNS.

Has anyone tried a tape method install ? It's on my todo list.

Tape install worked for me using tape.index listed in post #63. I didn't test if installer expects files in original order (=index needed) or not. Tape drive configuration is in GUI, no need for manual config file editing anymore. Tape drive must be SCSI unit #4.

Better do official tape installation first before attempting to solve possibly non-existing bugs due to hacked installer.

Yeah I'll give it a go once I've got a few more bits added to the toolchain. Thanks!

Tried the tape install both as the directory method (with an index.tape in it) and a tape.tar that contained the contents of the directory but not the directory itself as the tape drive... no dice :banghead . This was for a 2.1 install. Not sure what I'm doing wrong, any ideas?

No ideas without config file, logs and what exactly Amix says when it attempts to find the tape. Also does it see the tape drive in initial scsi detection?

I tested again, index is required, without index it only complains about errors.. (Note that index file is called index.tape, not original tape.index)

You should see something like this in the log:

Following comes after harddrive selection:

Index is wrong if first TAPEEMU: File 'xx' is not list.cpio and second isn't core.cpio.

Ok, it keeps failing when looking for file List... but the tape drive never seems to be detected despite being on SCSI4. Logs + screenshot attached.

EDIT: Can't see any TAPEEMU and I'm using WinUAE 2.7b6 :/ Maybe rename all the tape files to their "real" names with .cpio extension?

EDIT2: gotcha about the index.tape needing list.cpio and core.cpio as the two first entries in index.tape. Also winuaelog.txt DOES have TAPEEMU, sorry.

Ok I managed to get it to read the tape... but it didn't get very far once I told it to install every package on the tape. I think you have to rename each blob to blob.cpio according to index.tape's entries if you want anything to happen ... and there's a few of those cpio files that appear to be corrupted (I did a manual cpio of the various blobs within an already working AMIX 2.1p2a install to check). Hopefully it's the non critical stuff, like gnu sources. Here goes... Thanks again Toni!

EDIT: Confirmed, the install completes although some packages are definitely corrupted somehow. Unfortunately amigasrc seems to be among them... I'll see what I can do now I've succeeded in doing a tape install. This is SO retro and cool!