newest whdload not works on m68k

[olafsch1590-ctrl]

If I update whdload in aros 68k there is only the error "error in display initialisation could not find any UCL instruction" if I start a slave. The version before from end of 2025 worked but mostly after one or two whdload slaves were started and closed there was only a black screen or error requester with problem with "input device"

I have prepared a complete distribution based on nightly from 9.9. and whdload added including aros roms in devs:kickstarts

You can download it here: http://aros-vision.de/downloads/aros-amiga-m68k.zip

perhaps you could look at it. In sys:whdload there are lots of "whdload slaves" integrated for testing

winuae config as before from http://aros-vision.de/downloads/winuae.use

[olafsch1590-ctrl]

yesterday tested again. Not changed. Older version starts but only randomly works from time to time

[olafsch1590-ctrl]

to install older version just use it from this page https://whdload.de/whdload/old-whdload/ version from 28.10.2025 from one of the archives and copy directory C and S

[olafsch1590-ctrl]

that was not really solving the problem. The error is no longer there but then only black screen. No slave works. Amiblitz also not compiles. Tested today with newest nightly. Older whdload from last year starts slave but mostly only one time, if I start a second slave and close it it not returns to the desktop just black screen

[Kalamatee]

Thanks for the report. There were three separate things in it.

"Couldn't find UCL instruction" – this was an AROS bug in the display rethink path (RethinkDisplay/CMove) and is fixed in master (11d17ac). It is in the current nightly.

Crashes (illegal instruction / Line 1111 / address errors) in Wanderer right after WHDLoad returned – this was a real AROS bug. WHDLoad's exit flushes memory, which expunges unused libraries; on reload, AROS's per-task library bases (png.library, z1.library, posixc.library) could hand back a freed base from a previous incarnation, and the first call through it jumped into garbage. It is reproducible on every AROS port with Avail FLUSH between two uses of the PNG datatype and has nothing to do with WHDLoad. The fix is in genmodule (tools/genmodule/writestart.c) and every library gets rebuilt with it in the next nightly.

Black screen and no return to the desktop after quitting a slave – this is not an AROS bug and will not be worked on here. WHDLoad 20.0 changed how it switches copper/raster DMA off at exit (see its History.txt: "now performed via an user copper list"). Its exit code waits in a btst #2,INTREQR loop for a SOFTINT that its own restore copper list at chip $1004 is supposed to raise; when that list is cut short by the vertical blank restart, SOFTINT never comes and the machine sits on a black screen with interrupts disabled. I reproduced the identical hang, same loop and register state, with a real Kickstart 3.1 A1200 ROM in the same WinUAE setup, and WHDLoad 19.2 shows it too. Please raise it with the WHDLoad developers; the details above should be enough for them to find it.

Also note the files in Devs/Kickstarts in your archive are copies of the AROS ROM, not Kickstart images. Slaves that use KickEmu (VirtualWorld, A1200Demo and similar) will not work on any OS until real kick40068.A1200 / kick34005.A500 images are placed there.

[olafsch1590-ctrl]

I sent a email with the link to this page to one of the whdload developers. We will see what happens. Thanks for help