apple2 w/ SCSI card: need ability to control device 1

[inexorabletash]

In the apple2 driver (and mac, apparently?) the SCSI bus has ID 1 hard-coded to be a CD-ROM device and so options like -sl7:scsi:scsibus:1 harddisk are not supported. Can this be modified to allow specifying or removing the device by passing '' etc?

[inexorabletash]

cc: @ksherlock @arekkusu42

[ksherlock]

The config in question:

mame/src/devices/bus/a2bus/a2scsi.cpp

Lines 75 to 80 in 4ac6a50

	NSCSI_CONNECTOR(config, "scsibus:1").option_set("cdrom", NSCSI_CDROM_APPLE).machine_config(
	[](device_t *device)
	{
	device->subdevice<cdda_device>("cdda")->add_route(0, "^^speaker", 1.0, 0);
	device->subdevice<cdda_device>("cdda")->add_route(1, "^^speaker", 1.0, 1);
	});

mame/src/mame/apple/maclc.cpp

Lines 352 to 357 in 4ac6a50

	NSCSI_CONNECTOR(config, "scsi:3").option_set("cdrom", NSCSI_CDROM_APPLE).machine_config(
	[](device_t *device)
	{
	device->subdevice<cdda_device>("cdda")->add_route(0, "^^speaker", 1.0, 0);
	device->subdevice<cdda_device>("cdda")->add_route(1, "^^speaker", 1.0, 1);
	});

[rb6502]

I'm not actually sure why you can't remove or replace the device. That's odd.

[rb6502]

Actually, no, I get it. You can't have a freely configurable CD-ROM and working redbook audio. It's a thing right now. Is there some important use case to have something else on ID 1?

[galibert]

Errr, you can? machine_config applies to the output of option_add/set (aka the slot_option), not to the device_slot_interface. You could have a mac_scsi_devices_with_cdda_two_levels_up with the machine_config on the option_add of the cdrom.

Or maybe we can make the nscsi bus and the nscsi connector device_mixer_interfaces and have the nscsi cdrom always bring the audio up? And the connector bring it up too of course. Then systems in general will be able to pick up cdda (and any other analog audio feed, but I don't think there was anything else) from the bus. Which would you prefer?

[inexorabletash]

Is there some important use case to have something else on ID 1?

It's not critical.

My use case is that I'm building a test harness for Apple II DeskTop using MAME, where tests declare a configuration (slots, cards, etc) needed for the test - e.g. does it correctly navigate the weirdness of SmartPort, SCSI, and ProDOS each with their own notions of device locations. When a test configuration includes a SCSI card I run into the limitation that it must have a CD-ROM in ID1. Not much of a limitation, but it was a head-scratcher.

Here's a sample test file that goes way beyond the number of devices - https://github.com/a2stuff/a2d/blob/main/tests/desktop/format_erase_long_names.lua - this tries to load everything up with "aplcdsc" devices since those have the longest name ("SONYCD-ROMCDU-80") but a couple IDs are taken up by the "cdrom" devices. It's really an edge case, but that's what I'm testing.

[ajrhacker]

26da2f2 should fix this.

[rb6502]

This will be fixed in a less intrusive way. 26da2f2 was not it.

[ajrhacker]

Can you please explain?

[rb6502]

Errr, you can? machine_config applies to the output of option_add/set (aka the slot_option), not to the device_slot_interface. You could have a mac_scsi_devices_with_cdda_two_levels_up with the machine_config on the option_add of the cdrom.

Or maybe we can make the nscsi bus and the nscsi connector device_mixer_interfaces and have the nscsi cdrom always bring the audio up? And the connector bring it up too of course. Then systems in general will be able to pick up cdda (and any other analog audio feed, but I don't think there was anything else) from the bus. Which would you prefer?

I don't love either of those. The root problem is still that we have no framework for something that can implement more than one device_slot_interface at a time (SCSI + audio, or ATA + audio). Fortunately there's a better way in this specific case.