DOS ain't dead

I prepared a blog post at https://pushbx.org/ecm/dokuwiki/blog/pushbx/2024/0107_enhanced_dr-dos_single-file_load

Basically, I converted EDR-DOS to load from a single kernel file.

---
l

> I prepared a blog post at
> https://pushbx.org/ecm/dokuwiki/blog/pushbx/2024/0107_enhanced_dr-dos_single-file_load
>
> Basically, I converted EDR-DOS to load from a single kernel file.

cool, this is kind of DR-DOS 7.06.

> I prepared a blog post at
> https://pushbx.org/ecm/dokuwiki/blog/pushbx/2024/0107_enhanced_dr-dos_single-file_load
>
> Basically, I converted EDR-DOS to load from a single kernel file.

Good work. As your post stated, it open the doors to future improvements. Glad to see EDR is alive with you!

---
Visit my personal blog at https://www.javiergutierrezchamorro.com

> I prepared a blog post at
> https://pushbx.org/ecm/dokuwiki/blog/pushbx/2024/0107_enhanced_dr-dos_single-file_load
>
> Basically, I converted EDR-DOS to load from a single kernel file.

Great - both the changes and the blog post!
Can you please arrange for binary builds to be generated?

Does it support the LBA 33-bit?

Can you mention if some other features from lDOS, RxDOS, DR-DOS 7.03/8.x are not yet supported?

From your blog post I understand that December 2023 kernel can be booted from boot sectors of lDOS / RxDOS.3, FreeDOS, Enhanced DR-DOS, MS-DOS v6 / IBM-DOS, MS-DOS v7, Multiboot v1, Multiboot v2, NTLDR, BOOTMGR, or DOS-C IPL.SYS.

What about:
- booting from bootsectors of other DR-DOS (3.31-7.05), PC DOS 7.10
- boot sector that allows booting EDR-DOS, FreeDOS, PC DOS, MS-DOS v6, MS-DOS v7/8, DR-DOS
- boot sector that allows booting from GPT
- using COMMAND.COM from another DOS as main SHELL with the December 2023 kernel? (e.g. how much of these and these are implemented?) - the opposite question of this one.

> > I prepared a blog post at
> >
> https://pushbx.org/ecm/dokuwiki/blog/pushbx/2024/0107_enhanced_dr-dos_single-file_load
> >
> > Basically, I converted EDR-DOS to load from a single kernel file.
>
> Great - both the changes and the blog post!
> Can you please arrange for binary builds to be generated?

My repo already has automatic (daily) current builds. You can find the most recent current build in https://pushbx.org/ecm/download/edrdos.zip - This file will be updated to a 2024 January revision within the next 1 hour as of this posting. This revision includes most of the patches from the SvarDOS repository, excluding some for building with OpenWatcom make and wlink instead.

All four kernel files are found in the root of the zipball. The shell is in command/bin/command.com as before. The files drbio.bin and drdos.sys exist in the subdirectories but are used only to build the single-file kernels, not needed otherwise.

> Does it support the
> LBA 33-bit?

No, not yet.

> Can you mention if some other features from lDOS, RxDOS, DR-DOS 7.03/8.x
> are not yet supported?

RxDOS comes with an LFN server built in (corresponding to the DOSLFN tool), but this is very incomplete and not fully compatible.

> From your blog post I understand that December 2023 kernel

This is only true of the kernel files using iniload, the *.com files. The *.sys files only support loading as FreeDOS or EDR-DOS kernels.

> can be booted
> from boot sectors of lDOS / RxDOS.3, FreeDOS, Enhanced DR-DOS, MS-DOS v6 /
> IBM-DOS, MS-DOS v7, Multiboot v1, Multiboot v2, NTLDR, BOOTMGR, or DOS-C
> IPL.SYS.

Yes, but don't say the last ones too loudly.

> What about:
> - booting from bootsectors of other DR-DOS (3.31-7.05),

I would have to test and/or study these. If you can provide images that use them I can look into it.

> PC DOS 7.10

The IBM-DOS entry of lDOS iniload does support PC-DOS 7.10 load, including from FAT32.

> - boot sector that allows booting EDR-DOS, FreeDOS, PC DOS, MS-DOS v6,
> MS-DOS v7/8, DR-DOS

My ldosboot boot sector loaders can be built to cover a number of protocols, though you may need to disable some options to make them fit. It is easier to set up lDebug (using any sector-to-iniload protocol to load the debugger) then enter a boot protocol=PROTOCOLNAME command and upon success run a Q command to make the debugger hand over control to the next kernel.

> - boot sector that allows
> booting
> from GPT

You need another loader stage to replace the traditional MBR loader which will load and pass control to the boot sector loader. Other than the 32-bit or 33-bit LBA limits most boot sector loaders do not care whether they are running on a GPT, MBR primary, or MBR logical partition.

> - using COMMAND.COM from another DOS as main SHELL with the December 2023
> kernel? (e.g. how much of
> these
> and these are implemented?) -
> the opposite question of
> this one.

I mostly work on the kernels. If you have any particular feature of another shell that doesn't work on this kernel, feel free to specify.

---
l

> > - boot sector that allows booting EDR-DOS, FreeDOS, PC DOS, MS-DOS v6,
> > MS-DOS v7/8, DR-DOS
>
> My ldosboot boot sector loaders can be built to cover a number of
> protocols, though you may need to disable some options to make them fit. It
> is easier to set up lDebug (using any sector-to-iniload protocol to load
> the debugger) then enter a boot protocol=PROTOCOLNAME command
> and upon success run a Q command to make the debugger hand over control to
> the next kernel.

By the way, you can load a kernel file from any subdirectory of a (FAT12, FAT16, or FAT32) file system. For instance, boot protocol ldos hda1/kernels/edrdos.com could be used to load EDR-DOS from a subdirectory. Further, I fixed a bug recently which makes it so you can now load traditional dual-file kernels (MS-DOS v6 or IBM-DOS, but not old EDR-DOS) from any two subdirectories. boot protocol msdos6 fda/msdosio/msdos6io.sys /msdosdos/msdos6.sys would work. If the files are in the same directory, a command like boot protocol msdos6 fda/msdos/msdos6io.sys msdos6.sys works too (if the add name doesn't contain a slash it is in the same directory as the kernel name). Details in the manual.

---
l

Modified EDR-DOS cannot enter to one of the directories on my hard drive.

I tested kernel from 19 may 2024.

https://pushbx.org/ecm/download/old/edrdos/20240519.zip

> Modified EDR-DOS cannot enter to one of the directories on my hard drive.
>
> I tested kernel from 19 may 2024.
>
> https://pushbx.org/ecm/download/old/edrdos/20240519.zip

Does any earlier revision, including Udo Kuhnt's, work? Do other DOS versions work? What exactly does "cannot enter" mean, a CD command fails for this directory? Can you create a disk image that reproduces the problem in qemu? (Preferably with as little data on the drive as possible.)

---
l

> > Modified EDR-DOS cannot enter to one of the directories on my hard
> drive.
> >
> > I tested kernel from 19 may 2024.
> >
> > https://pushbx.org/ecm/download/old/edrdos/20240519.zip
>
> Does any earlier revision, including Udo Kuhnt's, work? Do other DOS
> versions work? What exactly does "cannot enter" mean, a CD
> command fails for this directory? Can you create a disk image that
> reproduces the problem in qemu? (Preferably with as little data on the
> drive as possible.)

The kernels of other clones work and so do previous versions of EDR-DOS (I tested until August 2023). The CD command does not work and neither does trying to enter the directory with e.g. Volkov Commander. I suspect that these are directories with a cluster larger than 65535. My disk is USB FAT32 - 32GB.

> The kernels of other clones work and so do previous versions of EDR-DOS (I
> tested until August 2023).

What exact revision is the latest that works?

> The CD command does not work and neither does
> trying to enter the directory with e.g. Volkov Commander. I suspect that
> these are directories with a cluster larger than 65535. My disk is USB
> FAT32 - 32GB.

I'll see about trying this later.

---
l

> > The kernels of other clones work and so do previous versions of EDR-DOS
> (I
> > tested until August 2023).
>
> What exact revision is the latest that works?

This could be cause for an error: https://hg.pushbx.org/ecm/edrdos/rev/8d3b6d9a415d

Does this one work?: https://pushbx.org/ecm/download/old/edrdos/20231205.zip

---
l

> > What exact revision is the latest that works?

Version <= 20240114.zip works fine
Version >= 20240329.zip doesn't work

> > > What exact revision is the latest that works?
>
> Version <= 20240114.zip works fine
> Version >= 20240329.zip doesn't work

I was able to reproduce, CD to a directory causes a critical error prompt. I built a hard disk image like this: nasm -D_MEDIAID=0F8h -D_SPF=2000 ../bootimg/bootimg.asm -I ../bootimg/ -I ../lmacros/ -D_MBR -D_BPE=32 -D_MBR_PART_TYPE=fat32 -o big32.img -D_PAYLOADFILE=::fill,'(512 * 76000),26h,fill.dat,::chdir,testdir,::fill,32,32,test.dat' -D_SPI=256000

I am fairly sure that a SvarDOS contribution changing the LBA flagging is to blame: https://hg.pushbx.org/ecm/edrdos/rev/220f54eecbb8 From https://github.com/SvarDOS/edrdos/commit/99765bd4c9ec9e035f1bd2af191c35d440c99383

It changed the LBA use to be indicated by the UDF_LBA flag. However, in add_unit this flag is cleared: https://hg.pushbx.org/ecm/edrdos/file/1137afe0115d/drbio/disk.asm#l616

The next changeset will fix this problem, please test today's June build in 45min from now.

---
l

> > > > What exact revision is the latest that works?
> >
> > Version <= 20240114.zip works fine
> > Version >= 20240329.zip doesn't work
>
> I was able to reproduce, CD to a directory causes a critical error prompt.
> I built a hard disk image like this: nasm -D_MEDIAID=0F8h -D_SPF=2000
> ../bootimg/bootimg.asm -I ../bootimg/ -I ../lmacros/ -D_MBR -D_BPE=32
> -D_MBR_PART_TYPE=fat32 -o big32.img -D_PAYLOADFILE=::fill,'(512 *
> 76000),26h,fill.dat,::chdir,testdir,::fill,32,32,test.dat'
> -D_SPI=256000
>
> I am fairly sure that a SvarDOS contribution changing the LBA flagging is
> to blame: https://hg.pushbx.org/ecm/edrdos/rev/220f54eecbb8 From
> https://github.com/SvarDOS/edrdos/commit/99765bd4c9ec9e035f1bd2af191c35d440c99383
>
> It changed the LBA use to be indicated by the UDF_LBA flag.
> However, in add_unit this flag is cleared:
> https://hg.pushbx.org/ecm/edrdos/file/1137afe0115d/drbio/disk.asm#l616
>
> The next changeset will fix this problem, please test today's June build in
> 45min from now.

This is the fix changeset: https://hg.pushbx.org/ecm/edrdos/rev/dde66e015fb8

---
l

> The next changeset will fix this problem, please test today's June build in
> 45min from now.

Version 20240607 works fine.

Could you create an alternative version of EDR-DOS and add detection of FAT drives on the GPT partition? See how this is done in FreeDOS.

Thanks.

InitDisk.C (FreeDOS)
https://megawrzuta.pl/download/078fe0c107fb750fb2f9120126d869e5.html

?

> Could you create an alternative version of EDR-DOS and add detection of FAT
> drives on the GPT partition? See how this is done in FreeDOS.

There are also a bunch of incompatibilities between FreeDOS / MS-DOS and EDR kernels regarding drive letter assignment
https://github.com/SvarDOS/edrdos/issues/59

and the permitted access to non-initialized FAT partitions
https://github.com/SvarDOS/edrdos/issues/15

If anyone is interested in working on this, feel free to join...

> ?

What is your question ?

---
--
http://glennmcc.org/

> > ?
>
> What is your question ?

His question is is in the post just above the interrogation point post

The DRBIO.SYS source is now fully ported to JWasm and can also be build under Win32 (and probably Linux if OpenWatcom WLINK is installed). Porting DRDOS.SYS will take some time. But after that the EDR-DOS kernel should be buildable with free software.

"STC" instructions are unnecessary in "dos7.asm" if there is a jump to error_exit() where the flags are changed and the interrupt is exited with the CF flag set.

        stc
        jmp     error_exit              ; no, then exit with error


This code may be added to EDR-DOS sources.

GPT_Signature           equ 00h
GPT_Revision            equ 08h
GPT_HeaderSize          equ 0Ch
GPT_HeaderCRC           equ 10h
GPT_Reserved            equ 14h
GPT_CurrentLBA          equ 18h
GPT_BackupLBA           equ 20h
GPT_FirstUsableLBA      equ 28h
GPT_LastUsableLBA       equ 30h
GPT_DiskGUID            equ 38h
GPT_PartitionsLBA       equ 48h
GPT_PartitionNumber     equ 50h
GPT_PartitionEntrySize  equ 54h

GPTEntry_PartitionGUID  equ 00h
GPTEntry_uniqGUID       equ 10h
GPTEntry_StartLBA       equ 20h
GPTEntry_EndLBA         equ 28h
GPTEntry_Attributes     equ 30h
GPTEntry_Name           equ 38h

        db      'GPT',0         ;gpt code flag

gpt_init:
        nop                     ;to modify by hex editor to "ret" - 0xC3
        mov     dl,80h          ;start from 1st hard disk
       ;push    di
ya_gpt:
        xor     di,di           ;\ start from zero partition
        xor     bx,bx           ;/

again_gpt:
        cmp     nunits,MAXPART          ; do we already have the maximum?
         jae    lt_gpt                  ; skip if space for more units

        lea     si,diskaddrpack         ; pointer to disk address packet
        mov     word ptr [si+ 8],1      ;\ make dword = 1
        and     word ptr [si+10],0      ;/
        push    bx
        call    login_read_dx_lba       ; read sector
        pop     bx
        jc      no_gpt
        mov     si,CG:local_buffer      ; check if sector start
        lodsw                           ; with "EFI PART" string
        cmp     ax,"FE"
        jnz     no_gpt
        lodsw
        cmp     ax," I"
        jnz     no_gpt
        lodsw
        cmp     ax,"AP"
        jnz     no_gpt
        lodsw
        cmp     ax,"TR"
        jnz     no_gpt

        ; start from, qword = 2?
        cmp     word ptr [si+GPT_PartitionsLBA-4*2+0],2
        jnz     no_gpt
        xor     ax,ax
        cmp     [si+GPT_PartitionsLBA-4*2+2],ax
        jnz     no_gpt
        cmp     [si+GPT_PartitionsLBA-4*2+4],ax
        jnz     no_gpt
        cmp     [si+GPT_PartitionsLBA-4*2+6],ax
        jnz     no_gpt

        ; partition size, qword = 80h?
        cmp     word ptr [si+GPT_PartitionEntrySize-4*2+0],0080h
        jnz     no_gpt
        cmp     word ptr [si+GPT_PartitionEntrySize-4*2+2],ax
        jnz     no_gpt

        ; save number of partitons
        push    word ptr [si+GPT_PartitionNumber-4*2+2]
        push    word ptr [si+GPT_PartitionNumber-4*2+0]

        ; save LBA sector number (dword)
        push    word ptr [si+GPT_PartitionsLBA-4*2+2]
        push    word ptr [si+GPT_PartitionsLBA-4*2+0]

        lea     si,diskaddrpack

        ; copy LBA sector number to packet and load this sector
        pop     word ptr [si+ 8]
        pop     word ptr [si+10]
        ; make partition number division by 4
        mov     ax,di
        mov     cx,bx
        shr     cx,1
        rcr     ax,1
        shr     cx,1
        rcr     ax,1
        ; add division result to the sector number
        add     word ptr [si+ 8],ax
        adc     word ptr [si+10],cx

        push    bx
        call    login_read_dx_lba       ; read sector
        pop     bx
        jc      er_gpt

        ; make pointer to 128 bytes structure in sector (512/128=4)
        mov     ax,di
        and     ax,3
        mov     cl,80h
        mul     cl
        add     ax,CG:local_buffer
        mov     si,ax

        ; load LBA disk starting sector number
        mov     ax,[si+GPTEntry_StartLBA+0*8+0]
        mov     cx,[si+GPTEntry_StartLBA+0*8+2]
        ; save partition start
        mov     word ptr [partstart+0],ax
        mov     word ptr [partstart+2],cx
        ; zero ?
        or      ax,cx
        jz      nx_gpt

        ; if partition finish (dword +4)<>0 then skip
        mov     ax,[si+GPTEntry_EndLBA+4+2]
        or      ax,[si+GPTEntry_EndLBA+4+0]
        jnz     nx_gpt

        ; save partition finish
        mov     ax,[si+GPTEntry_EndLBA+2]
        mov     word ptr [partend+2],ax
        mov     ax,[si+GPTEntry_EndLBA+0]
        mov     word ptr [partend+0],ax

        mov     ax,word ptr [partstart+0]
        mov     cx,word ptr [partstart+2]

        ; compute partition size
        mov     si,word ptr [partend+0]
        sub     si,ax
        mov     word ptr [part_size+0],si
        mov     si,word ptr [partend+2]
        sbb     si,cx
        mov     word ptr [part_size+2],si
        add     word ptr [part_size+0],1
        adc     word ptr [part_size+2],0

        lea     si,diskaddrpack         ; pointer to disk address packet
        mov     word ptr [si+ 8],ax
        mov     word ptr [si+10],cx
        push    bx
        call    login_read_dx_lba       ; read 1st partition sector
        pop     bx
        jc      nx_gpt

        mov     si,CG:local_buffer
        mov     ax,[si+52h]             ;\
        cmp     ax,"AF"                 ; \
        jnz     nx_gpt                  ;  > = "FAT"? to skip NTFS
        mov     al,[si+54h]             ; /
        cmp     al,"T"                  ;/
        jnz     nx_gpt

        mov     [parttype],FAT32X_ID

        pushx   <di,bx,dx>
        call    login_p0        ; add and register new disk
        popx    <dx,bx,di>
nx_gpt:
        add     di,1            ;\ next partition
        adc     bx,0            ;/

        pop     ax              ;\ left partitions
        pop     cx              ;/

        cmp     cx,bx
        jnz     again_gpt
        cmp     ax,di
        jnz     again_gpt

        jmp     no_gpt
er_gpt:
        add     sp,2*2
no_gpt:
        inc     dl              ; next hard disk to check
        jnz     ya_gpt          ; if ZF=0 jump
lt_gpt:
       ;pop     di
        ret


> This code may be added to EDR-DOS sources.

As I said: just 187 assembly lines (including empty lines).

more like 50 lines in C.

What is missing is the part where partition start and ending offsets are beyond
2^32 (almost trivial) AND WHERE TO STORE these offsets, adding them to LBA48 offsets before passing to the BIOS.

> This code may be added to EDR-DOS sources.

As I said: just 187 assembly lines (including empty lines).

probably more like 50 lines in C.

What is missing is the part where partition start and ending offsets are
beyond 2^32 (almost trivial) AND WHERE TO STORE these offsets, adding them to
LBA48 offsets before passing them to the BIOS.

> > This code may be added to EDR-DOS sources.
>
> As I said: just 187 assembly lines (including empty lines).
>
> probably more like 50 lines in C.
>
> What is missing is the part where partition start and ending offsets are
> beyond 2^32 (almost trivial) AND WHERE TO STORE these offsets, adding them
> to
> LBA48 offsets before passing them to the BIOS.

I made some small corrections. Thanks.

GPT_Signature           equ 00h
GPT_Revision            equ 08h
GPT_HeaderSize          equ 0Ch
GPT_HeaderCRC           equ 10h
GPT_Reserved            equ 14h
GPT_CurrentLBA          equ 18h
GPT_BackupLBA           equ 20h
GPT_FirstUsableLBA      equ 28h
GPT_LastUsableLBA       equ 30h
GPT_DiskGUID            equ 38h
GPT_PartitionsLBA       equ 48h
GPT_PartitionNumber     equ 50h
GPT_PartitionEntrySize  equ 54h

GPTEntry_PartitionGUID  equ 00h
GPTEntry_uniqGUID       equ 10h
GPTEntry_StartLBA       equ 20h
GPTEntry_EndLBA         equ 28h
GPTEntry_Attributes     equ 30h
GPTEntry_Name           equ 38h

        db      'GPT',0         ;gpt code flag

gpt_init:
        nop                     ;to modify by hex editor to "ret" - 0xC3
        mov     dl,80h          ;start from 1st hard disk
       ;push    di
ya_gpt:
        xor     di,di           ;\ start from zero partition
        xor     bx,bx           ;/

again_gpt:
        cmp     nunits,MAXPART          ; do we already have the maximum?
         jae    lt_gpt                  ; skip if space for more units

        lea     si,diskaddrpack         ; pointer to disk address packet
        mov     word ptr [si+ 8],1      ;\ make dword = 1
        and     word ptr [si+10],0      ;/
        push    bx
        call    login_read_dx_lba       ; read sector
        pop     bx
        jc      no_gpt
        mov     si,CG:local_buffer      ; check if sector start
        lodsw                           ; with "EFI PART" string
        cmp     ax,"FE"
        jnz     no_gpt
        lodsw
        cmp     ax," I"
        jnz     no_gpt
        lodsw
        cmp     ax,"AP"
        jnz     no_gpt
        lodsw
        cmp     ax,"TR"
        jnz     no_gpt

        ; start from, qword = 2?
        cmp     word ptr [si+GPT_PartitionsLBA-4*2+0],2
        jnz     no_gpt
        xor     ax,ax
        cmp     [si+GPT_PartitionsLBA-4*2+2],ax
        jnz     no_gpt
        cmp     [si+GPT_PartitionsLBA-4*2+4],ax
        jnz     no_gpt
        cmp     [si+GPT_PartitionsLBA-4*2+6],ax
        jnz     no_gpt

        ; partition size, qword = 80h?
        cmp     word ptr [si+GPT_PartitionEntrySize-4*2+0],0080h
        jnz     no_gpt
        cmp     word ptr [si+GPT_PartitionEntrySize-4*2+2],ax
        jnz     no_gpt

        ; save number of partitons
        push    word ptr [si+GPT_PartitionNumber-4*2+2]
        push    word ptr [si+GPT_PartitionNumber-4*2+0]

        ; save LBA sector number (dword)
        push    word ptr [si+GPT_PartitionsLBA-4*2+2]
        push    word ptr [si+GPT_PartitionsLBA-4*2+0]

        lea     si,diskaddrpack

        ; copy LBA sector number to packet and load this sector
        pop     word ptr [si+ 8]
        pop     word ptr [si+10]
        ; make partition number division by 4
        mov     ax,di
        mov     cx,bx
        shr     cx,1
        rcr     ax,1
        shr     cx,1
        rcr     ax,1
        ; add division result to the sector number
        add     word ptr [si+ 8],ax
        adc     word ptr [si+10],cx

        push    bx
        call    login_read_dx_lba       ; read sector
        pop     bx
        jc      er_gpt

        ; make pointer to 128 bytes structure in sector (512/128=4)
        mov     ax,di
        and     ax,3
        mov     cl,80h
        mul     cl
        add     ax,CG:local_buffer
        mov     si,ax

        ; StartLBA >= 2^32
        mov     ax,[si+GPTEntry_StartLBA+4+0]
        or      ax,[si+GPTEntry_StartLBA+4+2]
        jnz     nx_gpt

        ; load LBA disk starting sector number
        mov     ax,[si+GPTEntry_StartLBA+0*8+0]
        mov     cx,[si+GPTEntry_StartLBA+0*8+2]
        ; save partition start
        mov     word ptr [partstart+0],ax
        mov     word ptr [partstart+2],cx
        ; zero ?
        or      ax,cx
        jz      nx_gpt

        ; if partition finish (dword +4)<>0 then skip (EndLBA >= 2^32)
        mov     ax,[si+GPTEntry_EndLBA+4+2]
        or      ax,[si+GPTEntry_EndLBA+4+0]
        jnz     nx_gpt

        ; save partition finish
        mov     ax,[si+GPTEntry_EndLBA+2]
        mov     word ptr [partend+2],ax
        mov     ax,[si+GPTEntry_EndLBA+0]
        mov     word ptr [partend+0],ax

        mov     ax,word ptr [partstart+0]
        mov     cx,word ptr [partstart+2]

        ; compute partition size
        mov     si,word ptr [partend+0]
        sub     si,ax
        mov     word ptr [part_size+0],si
        mov     si,word ptr [partend+2]
        sbb     si,cx
        jc      nx_gpt                  ; overflow? yes, skip
        mov     word ptr [part_size+2],si
        add     word ptr [part_size+0],1
        adc     word ptr [part_size+2],0

        lea     si,diskaddrpack         ; pointer to disk address packet
        mov     word ptr [si+ 8],ax
        mov     word ptr [si+10],cx
        push    bx
        call    login_read_dx_lba       ; read 1st partition sector
        pop     bx
        jc      nx_gpt

        mov     si,CG:local_buffer
        mov     ax,[si+52h]             ;\
        cmp     ax,"AF"                 ; \
        jnz     nx_gpt                  ;  > = "FAT"? to skip NTFS
        mov     al,[si+54h]             ; /
        cmp     al,"T"                  ;/
        jnz     nx_gpt

        mov     [parttype],FAT32X_ID

        pushx   <di,bx,dx>
        call    login_p0        ; add and register new disk
        popx    <dx,bx,di>
nx_gpt:
        add     di,1            ;\ next partition
        adc     bx,0            ;/

        pop     ax              ;\ left partitions
        pop     cx              ;/

        cmp     cx,bx
        jnz     again_gpt
        cmp     ax,di
        jnz     again_gpt

        jmp     no_gpt
er_gpt:
        add     sp,2*2
no_gpt:
        inc     dl              ; next hard disk to check
        jnz     ya_gpt          ; if ZF=0 jump
lt_gpt:
       ;pop     di
        ret


I finished porting the EDR-DOS kernel[1] to JWasm and was able to build it under DOS, Win32 and MacOS. This should also build under Linux. OpenWatcom, specifically wmake, wlink and exe2bin are additional build requirements for building drbio.sys and drdos.sys. The command.com build also depends on the Watcom C compiler.

Porting required some manual steps, opening the door for human errors. Time will tell how many of them I added to the source...

[1] https://github.com/SvarDOS/edrdos

Binaries and a 1.44M floppy image are now automatically generated via Github workflows. They can be downloaded from https://github.com/SvarDOS/edrdos/actions. Click on the last successful workflow run to get to the files...

> Binaries and a 1.44M floppy image are now automatically generated via
> Github workflows. They can be downloaded from
> https://github.com/SvarDOS/edrdos/actions. Click on the last successful
> workflow run to get to the files...

https://github.com/SvarDOS/edrdos/actions

---
MS-DOS forever!

> I finished porting the EDR-DOS kernel[1] to JWasm and was able to build it
> under DOS, Win32 and MacOS. This should also build under Linux. OpenWatcom,
> specifically wmake, wlink and exe2bin are additional build requirements for
> building drbio.sys and drdos.sys. The command.com build also depends on the
> Watcom C compiler.
>
> Porting required some manual steps, opening the door for human errors. Time
> will tell how many of them I added to the source...
>
> [1] https://github.com/SvarDOS/edrdos

I picked most of the changes to port DRBIO into my EDR-DOS repo. I used my new identicalising tool ident86 to validate that all changes in the binary are due to different encoding choices that have the same meaning in the disassembly. I wrote about this on my blog.

It turns out your DRBIO port seems to be completely right, except for a few problems along the way like the initcode: WRT keyword you used in this commit for nlsfunc.asm which you silently undid in a commit described as "convert bdosldr.a86 to JWASM". That's not good. I also made sure in my picks to separately copy, move, rename, or delete files in separate changesets that do not change any of the content in the same changeset, to simplify following the changes if you were to inspect the revisions' logs.

I didn't get around to applying the same process to the DRDOS module yet. As mentioned, this may require a small adjustment to ident86 to work with the DRDOS binary and its associated trace listing file drdos.tls created by the scripts in my build.

---
l

> It turns out your DRBIO port seems to be completely right, except for a few
> problems along the way like the initcode: WRT keyword you used
> in
> this commit for nlsfunc.asm which you silently undid
> in
> a commit described as "convert bdosldr.a86 to JWASM". That's not
> good. I also made sure in my picks to separately copy, move, rename, or
> delete files in separate changesets that do not change any of the content
> in the same changeset, to simplify following the changes if you were to
> inspect the revisions' logs.

Really helpful having someone double checking this! Thanks.

I learned a couple of things. Especially what not to do when converting a file: Copy it to a new file, convert the new file, check in the new file, and delete the old file. Good reciepe to make tracking the changes via commit log really hard. Better way is to commit as the original file and then rename. So you get a more meaningful diff. I tried to follow this later when porting DRDOS component.

>
> I didn't get around to applying the same process to the DRDOS module yet.
> As mentioned, this may require a small adjustment to ident86 to work with
> the DRDOS binary and its associated trace listing file drdos.tls created by
> the scripts in my build.

In contrast to the port of DRBIO, I tried not to introduce offset changes resulting from different instruction encodings, inserting NOPs etc. if needed. Made life easier when doing binary comparisons.

It would be useful to change some instructions "LEA Dst,Addr" to the shorter "MOV Dst,OFFSET Addr"

> It would be useful to change some instructions "LEA Dst,Addr" to the
> shorter "MOV Dst,OFFSET Addr"

Yes, I can give you access to the repo at https://github.com/SvarDOS/edrdos

> It would be useful to change some instructions "LEA Dst,Addr" to the
> shorter "MOV Dst,OFFSET Addr"

Apparently this is done by some version of MASM, but JWasm doesn't do it.

---
l

> > It would be useful to change some instructions "LEA Dst,Addr" to the
> > shorter "MOV Dst,OFFSET Addr"
>
> Apparently this is done by some version of MASM, but JWasm doesn't do it.

TASM and A86 do it by default, but newer MASMs (since v5?) don't do it anymore. Some purists complain when an assembler does things "behind your back".

IIRC, the recommended fix is via a macro with opattr 8 (or such). You probably have to use "option nokeyword:<lea>" first if you insist on calling the macro LEA.

It only saves a byte per instance and probably isn't noticeably faster.

> MASM 6.11 README.TXT
> The MASM 6.1 Reference indicates that the LEA instruction is
> encoded as a MOV when the source operand is a direct memory address.
>
> In response to programmer requests, MASM 6.1x no longer performs this
> optimization automatically. The optimization can be performed by
> using the OPATTR operator, as shown in the following macro:

          MOVLEA MACRO Dest, Symbol
          IF (OPATTR(Symbol)) AND 08h
               MOV Dest, OFFSET Symbol
          ELSE
               LEA Dest, Symbol
          ENDIF
          ENDM