2025-01-26

A user reported that PC-DOS command.com (after hacking its version check) didn't work correctly running a REN command on EDR-DOS. I was able to reproduce the error, and found that it was present in OpenDOS v7.01 already (but not in DR-DOS v6.00 or Novell DOS v7.00). The bug was that the destination of an FCB rename was stored to the wrong pointer variable, which wasn't helped by the source texts' use of hardcoded offsets from an FDOS parameter block. I described the error in some detail in the changeset message.

I also updated the lDOS version. Rather than "7.01.08 WIP (Month)" it reads "7.01.08 lDOS (Month)" now.

Update DM Extension for lDebug (ELD) that provides the DM command to additionally support listing of SD sub-MCBs. This is by default only enabled for the ELD, albeit the build option works for the main debugger too. The SKIPSD keyword reverts to the old behaviour. The manual was updated to describe the addition as well.

• Address most DOSENTRY interrupt entrypoints using segment 26h. This is a purely cosmetic change. The common lDOS entrypoint for relocated entries already pushed the original CS for reference, then did a far jump to normalise the segment.
• Enable BUF2 define to use adapted MS-DOS v2.11 buffer subsystem.
• Check array bounds in multDOS functions 2Ah and 2Eh.
• Split dosseg.nas file into ddataseg.nas and dcodeseg.nas to allow placing other sections between them when needed. (That is, for msbio1.nas.)
• In int 21h function 53h detect sector size of zero and handle it in a special way. I came across this while trying to boot lMS-DOS in dosemu2. Ultimately this change isn't needed because the zero sector size likely was from an uninitialised BPB due to a missing UPB.
• In the sysinit1.nas STALL loop, call int3 for debuggers and run HLT to idle the machine. A direct jump to itself was bad because dosemu2 detected it and terminated itself. This even seems to have lost the last message written to the DOS terminal.
• Allow to log in one primary partition per hard disk unit, no matter how many units there are. Previously only the first two units (80h and 81h) would be considered. Logical partitions handling unchanged. This allows to boot from the 2019 dosemu2 revision on our server.
• Optimise call 5 entry, after it was mentioned in a dosemu2 issue.
• Run the early DOSCODE segment from the tail of the sysinit segment, relocate it to the low end of the LMA later.
• Remove some obsolete parts of msdos/msinit.nas
• Put i00 entry into DOSENTRY segment, which transfers to DOSCODE then to DOSDATA
• Same for casemap
• Allow relocate and relocated mmacro for the same entrypoint within a single assembly unit, working around a NASM bug
• Add memory.asm from lDOS repo
• Adapt lDOS memory.asm
• Fix difficult to debug error due to not initialising all of the initial PSP (after the change to place it directly where wanted). Details in the changeset message.
• Update attribution lines in lDOS files
• Initialise the second byte of the int 31h vector (to zero)
• Support an int 31h entrypoint similar to RxDOS v7.24, transferring from DOSDATA:0 to the end of DOSDATA, then to DOSENTRY, then to DOSCODE. It returns an error unconditionally for now.
• Make place for and initialise a first UMCB at the end of the LMA.
• Support multDOS function 61h (from PTS-DOS) to return address of the first UMCB. Because there is such a large gap between the last function in the dispatch table (2Fh) and this one, we added a special check for it instead of adding it to the table.
• Support get and set UMB link status functions 5802h and 5803h. Also support extended strategy (high byte) for 5800h and 5801h.
• Add init.asm from lDOS repo.
• Include init.asm from sysinit1.nas, and enable building it.
• Harden the K09 check once. (I also asked over at retrocomputing stackexchange what the K09 is.)
• In init.asm doubley comment out unused section directives which confused convlist.pl
• Call lDOS init relocate functions (relocate device after every DEVICE= line, relocate end at the end). This chiefly includes the allocation and linking of UMBs into our UMCB chain, making the UMBs usable for programs that know to ask for them. Also, DOSCODE may be relocated into an UMB now.
• Use lDOS memory function to allocate largest block from two areas in exec.nas, which allows to use LH (eg in FreeCOM) to load programs into the UMA even if the largest LMA block is larger. This is not solved entirely yet, as flat-format (.COM) programs should have their size checked before allocating memory. As is, a small UMB may be allocated despite not being large enough for the program.
• Improve K09 check so older dosemu2 isn't misdetected as one. As discussed over at the dosemu2 repo.