This is with Debian's "DOSBox version 0.74-3".
HDPMI32.EXE is from https://github.com/Baron-von-Riedesel/HX/releases/tag/v2.19pre5
The test files can be found in https://ulukai.org/ecm/test/20210127.zip or can be built from https://hg.ulukai.org/ecm/ldebug and https://hg.ulukai.org/ecm/dpmitest
Now enter the following from the directory containing these files:
dosbox
keyb gr
mount C: .
C:
hdpmi32 -r
ldebugx.com dpmimini.com
g
g 16C
t
Expected: Trace step instruction at 0000_016Ch and register dump of next instruction (at 0000_0172h) while still in Protected Mode.
Actual: Execution ends up so that the dump occurs at CS:IP = 01A2h:02EAh in Real 86 Mode, which appears to be in some handler of HDPMI. Subsequent "G" commands to the debugger end up giving "Unexpected single-step interrupt" messages and dumping at the same CS:IP, though register values change and it appears every 5 runs one letter of the "Welcome in 32-bit protected mode" message is displayed.
Attaching the pre-release archive with the HDPMI32.EXE that I used.
Unfortunately the sourcecode of the extender is not available.
As after some debugging, we aren't 100% if we are doing something wrong. (a different pair of eyes might see something else though)
the missing behaviour has been discovered (e.g. what the extender wants).
Can you give 4426 a try ?
Is there some way to obtain a build of that or do I have to build it by my self?
On linux, it might be easiest to compile yourself.
you can use http://source.dosbox.com/dosboxsvn.tgz
I believe this SVN revision was included in the DOSBox-X 0.83.10 release, and it works fine as I wrote in their report: https://github.com/joncampbell123/dosbox-x/issues/2221#issuecomment-770781334 So thank you!
https://github.com/Baron-von-Riedesel/HX/tree/6bf77b09814e066abd0576cda88689852d1d13fe/Src/HDPMI This is the HDPMI sources.
Apologies, I accidentally put a symlink into the archive instead of the dpmimini.com executable. Fixed in this attachment.
Also reported at https://github.com/joncampbell123/dosbox-x/issues/2221