Hi
The config.sys parser is probably complaining that pmmerge.dll - which would be required to run pmshell.exe - is incorrect and/or damaged. Which may be a clue...
If Martin can boot from a different installation on that system (or install cd) it may be worth checking pmmerge.dll actually exists and looks correct - size and date wise - but I would start with my next suggestion.
Run chkdsk against the boot drive and then see if it can then boot OK.
Do *NOT* be fooled by a line like this in the config.sys file:-
IFS=L:\OS2\JFS.IFS /LW:5,20,4 /AUTOCHECK:*
What the above line does is simply check whether the flag indicating filesystem status is "clean"; it does not perform a chkdsk unless the flag says "dirty". If the system has not shutdown properly then the filesystem can be "dirty" but that flag may indicate "clean". That would certainly cause problems like Martin is experiencing.
As acpi is in such an unreliable state I now use this line in config.sys to check drives fully at each reboot - been caught too many times with inaccessible drives despite a supposedly "clean" shutdown when acpi.psd is in use:-
IFS=L:\OS2\JFS.IFS /LW:5,20,4 /AUTOCHECK:+G+H+I+J+K+L
Also worth mentioning: RC7 seems to "mung" ini files quickly compared to any other install of oS/2 I've ever had - dating back to 1995 - but I have no idea why. Maybe it is just something on my system but I run checkini and cleanini weekly on that system as opposed to about every 3 months on RC6a and 1.2 systems - same apps installed, same box, different boot volumes. Doubt if it is related to the current problem but...
Further advice for Martin when system is up'n'running:-
Check your acpi.psd line; you may benefit from using /SMP and /APIC plus whatever other switches the acpi docs indicate are suitable for Intel chipsets. You may also benefit from a later or earlier build of acpi - 3.16 is absolutely useless on nVidia nForce4 and 430 chipsets; I guess it may work fine on the Intel 82801...
Uniaud114RC6 is not very HDA capable and you may find the Paul Smedley builds work better - especially when you get acpi working fully. I guess you should start with the latest and work backwards
http://download.smedley.info/uniaud-1.9.22-release.zipPrevious builds all start with uniaud-1.9. OTOH: I'm sure I saw something about an Intel HDA specific build somewhere... Login to the "restricted betazone" via Mensys and you will find a uniuadtst4.zip (Yes, that is the spelling) package.
I hope the above chkdsk suggestion cures the problem.
Pete