Page 5 of 18
Re: QLiberator decompiler
Posted: Mon Dec 30, 2019 8:26 pm
by RWAP
Yes, this is some excellent progress - and the decompilers provide great tools for future preservation projects
Re: QLiberator decompiler
Posted: Tue Dec 31, 2019 11:07 am
by Martin_Head
EmmBee wrote:The manual says: "The minimum size of a channel table is 3 entries, for channels 0, 1 and 2" - so its the actual number of channels that need to be given.
CMD$ is initially set up by QLiberator as a variable to read the incoming Option command string and can be used as a normal variable afterwards.
Sounds like you're making excellent progress. Are you compiling with QDOS or SMSQ/E?
In the source code I have a 'REMark $$chan=9' But after compiling, at the start of the program, see the code snippet above, the number of channels should be 9, but I get 10.
The decompiler assigned CMD$ it's own variable name. But after recompiling, QLiberator got upset about trying to read a string (the wrongly named CMD$) that should have something in it.
I wrote and compiled all my test programs for the decompiler in QDOS under Qemulator. But I have started using QPC2 to recompile QLiberator and the Patch program, due to the amount of time it takes in Qemulator. It takes minutes just to load the QLIB_BAS program, and more minutes to compile it. Whereas in QPC2 the compile time is like 1 second with the BASIC compiler compiling itself, and 0 seconds with the original Qliberator. When using a _WRK file.
Re: QLiberator decompiler
Posted: Tue Dec 31, 2019 11:37 am
by EmmBee
I am using QLiberator version 3.36.
When I compile, in the top right-hand corner I get the message version 3.35
I've just tried compiling a program with a REMark $$chan=9
and QLiberator reports: Highest channel : 9
Re: QLiberator decompiler
Posted: Tue Dec 31, 2019 11:51 am
by Martin_Head
EmmBee wrote:I am using QLiberator version 3.36.
When I compile, in the top right-hand corner I get the message version 3.35
I've just tried compiling a program with a REMark $$chan=9
and QLiberator reports: Highest channel : 9
What does the QLIB_PATCH program report as the number of channels, or if you disassemble the start of the program and look at the stored number of channels.
I think I see the the same as you describe, QLiberator reports 9 channels, but 10 is actually stored in the program.
Also I'm pretty sure (I'm not sitting at the right computer at the moment), My Qliberator reports it's version as 3.36. And the PQLIB_PATCH program is 3.35
Re: QLiberator decompiler
Posted: Tue Dec 31, 2019 12:40 pm
by pjw
Im using V3.36. It reports V3.36 during compilation. QLib_obj has 9 channels stored in location $32. Other compiled programs appear to have $$chan=n + 1 channels stored in the program.
Re: QLiberator decompiler
Posted: Tue Dec 31, 2019 1:35 pm
by RalfR
Martin_Head wrote:There is one other odd thing I have noticed when comparing the original QLIB with my recompiled one's. There is a bit of data stored between the end of name list, and the start of the name/variable table
Perhaps this is something with the mouse handling. Marcel has cleaned one of the extensions in QLib for me (The extension for the mouse things) and they just work, if a bit of code is set in the heap, perhaps they store the mouse coordinate there.
Marcel may better clarify this.
Re: QLiberator decompiler
Posted: Tue Dec 31, 2019 5:11 pm
by Martin_Head
I think I have figured out what the 'something odd' is. It's 'External Procedures and Functions'. In the QLiberator manual around page 14.9 it talks about using QLIB! to start the compiler from BASIC. I've just got to figure out exactly how it works, to get the decompiler to recognize it.
I think the description of $$chan= , in the manual. Should say something like, It's the highest channel number used by the program. That gives you the extra 1 for channel 0.
QLiberator uses channels 3,4,5,6, and 7. So if you specify $$chan=8, then you either leave one spare channel, or maybe one of the built in SuperBASIC extensions uses a channel.
Looking in the source for QLIB_BAS I find
Code: Select all
1850 var04E8$ = "02-NOLINE,03-STAT,04-NODIS,05-NONAMES,06-AUTOF,07-RUN,08-DEBUG,09-NOBEEP,10-NOWINDS,11-TRACE,20-OBJ,21-LIST,23-NAME,"
which looks like the list of instructions for CMD$.
They don't quite match my QLiberator manual. -WINDS is missing. But -NOBEEP, -NOWINDS, and -TRACE are fair enough. But what's -NODIS. No display??
Anyone fancy playing with it, to find out what it does?
Re: QLiberator decompiler
Posted: Tue Dec 31, 2019 5:38 pm
by pjw
The first time I noticed there was a "no display" option was with some software that came from JMS. It was a Thing allowing you to compile a program directly from QD. Its a long time ago, and I didnt use it for real, so no details. JMS may have patched QLib? It seemed so at the time..
Re: QLiberator decompiler
Posted: Fri Jan 03, 2020 9:57 pm
by mk79
I have most of this stuff already figured out, but you're too fast for me
i'm too busy with other stuff right now, but here's the start of QLIB_obj as I currently understand it
Code: Select all
base: bra.w qlib_start
; ---------------------------------------------------------------------------
dc.w 0
dc.w $4AFB
dc.w 9
dc.b 'Qlib_3.35'
dc.b ' '
dc.l token_base-base
dc.l name_table-base
dc.l 0 ; Line table
dc.w 4
dc.l data_tokens-base ; First DATA token
dc.w 9 ; Channel count
dc.w $80 ; Buffer size
dc.w 512 ; Heap chunk size
dc.l 7000 ; Heap size
dc.w 916 ; Return stack size
dc.w 816 ; Name table entry of CMD$ variable
dc.l name_list-base ; Name list
dc.l ext_qutil-base ; Extensions
dc.l 20386 ; Data space size
dc.l 5500 ; Stack size
dc.b 0 ; STATS
dc.b 1 ; AUTOF
dc.l externals-base ; External procedures
dc.w 1
dc.w $FF00 ; WINDS
dc.l 0
dc.l 0
dc.l 0
dc.l 0
dc.l 0
dc.l 0
dc.l 0
dc.l 0
dc.l 0
dc.l 0
dc.w 0
; ---------------------------------------------------------------------------
job_start:
bra.w job_entry
; ---------------------------------------------------------------------------
external_entry:
bra.w external_entry_2
job_entry:
moveq #sms.info,d0
trap #1
lea base(pc),a1
cmpa.l a1,a6
beq.s job_do ; Normal jobs have a1 = a6
cmpa.l 2(a6),a1 ; Started by JMP trampoline?
bne.s init_externals ; LRESPRed extension, just initialise externals
job_do:
moveq #0,d6
suba.l a4,a5
move.l a5,d4
adda.l a6,a4
movea.l a1,a5
bsr.w check_embedded_rte
beq.s job_rte_init
movea.l sys_sbab(a0),a0 ; 'QL SuperBASIC' Area Base
adda.w #sb_offs,a0 ; offset from JCB to SuperBASIC vars
move.l sbq_runtime(a0),d0 ; System RTE?
beq.s no_system_runtime
movea.l d0,a0
job_rte_init: jsr 8(a0) ; Main runtime entry
job_exit: moveq #-1,d1
moveq #sms.frjb,d0 ; Forced Remove JoB
trap #1
no_system_runtime:
bsr.s write_rte_missing
bra.s job_exit
; End of function base
; =============== S U B R O U T I N E =======================================
write_rte_missing:
lea txt_runtime_missing(pc),a1 ; "\nRuntimes missing !\n"
suba.l a0,a0
movea.w (ut.wtext).w,a2 ; Write TEXT
jsr (a2)
rts
; End of function write_rte_missing
; =============== S U B R O U T I N E =======================================
check_embedded_rte:
lea embed_rte_base(pc),a1
cmpi.w #$6000,(a1) ; bra.l
bne.s no_embedded_rte
cmpi.w #$6000,4(a1) ; bra.l
bne.s no_embedded_rte
movea.l a1,a0
no_embedded_rte:rts
; End of function check_embedded_rte
; =============== S U B R O U T I N E =======================================
; Token address of external in d4
externals_entry:
lea base(pc),a4
bsr.s check_embedded_rte
beq.s jmp_embedded_rte
move.l sbq_runtime(a6),d0 ; Try to get system RTE
beq.s rte_not_found
movea.l d0,a0
jmp_embedded_rte:
jmp $10(a0)
; ---------------------------------------------------------------------------
rte_not_found:
bsr.s write_rte_missing
bra.s exit
; ---------------------------------------------------------------------------
init_externals:
movea.l a1,a4
move.l $52(a4),d4 ; External proc/fn list
bra.s init_externals_start
; ---------------------------------------------------------------------------
init_exernals_loop:
move.l (a5),d4
init_externals_start:
beq.s exit
lea (a4,d4.l),a5
movea.l a5,a1
addq.l #4,a1
move.l a4,-(sp)
movea.w (sb.inipr).w,a2 ; INITialise PRocedure table
jsr (a2)
movea.l (sp)+,a4
bra.s init_exernals_loop
; ---------------------------------------------------------------------------
exit: moveq #0,d0
rts
; End of function externals_entry
; ---------------------------------------------------------------------------
txt_runtime_missing:
dc.w 20
dc.b $A
dc.b 'Runtimes missing !',$A
I have disassembled most of the runtime, the other libraries of qlib_sys and the private QLIB_obj extensions. And with "disassembled" I mean the code is probably better commented than the original source code ever was
But it's still so much work to do and I'm at a point where it frankly gets hard to justify doing this just for fun. Maybe I need to start a gofundme campaign or something like that
Cheers, Marcel
Re: QLiberator decompiler
Posted: Sat Jan 04, 2020 11:10 am
by Martin_Head
mk79 wrote:I have most of this stuff already figured out, but you're too fast for me
i'm too busy with other stuff right now, but here's the start of QLIB_obj as I currently understand it
Code: Select all
dc.l 0 ; Line table
dc.w 4
dc.l data_tokens-base ; First DATA token
dc.w $FF00 ; WINDS
I have disassembled most of the runtime, the other libraries of qlib_sys and the private QLIB_obj extensions. And with "disassembled" I mean the code is probably better commented than the original source code ever was
But it's still so much work to do and I'm at a point where it frankly gets hard to justify doing this just for fun. Maybe I need to start a gofundme campaign or something like that
Cheers, Marcel
You've identified a few more things in that header than I have.
The dc.w 4 just after the line number table. I don't know exactly what it does, but in the QLIB source, if it's 0,1, or greater than 4, you get a BEEP, and maybe something else, otherwise it reads some value.
The 'dc.w $FF00 ; WINDS'. If I remember rightly, the QLIB source reads it as a byte, rather than a word, and treats it as a true/false flag.
You say you have done the private QLIB_obj extensions, Do you mean the embedded SuperBASIC extensions, with all the pointer interface and GENCODE type extensions? As I was thinking about doing them. Because I thought you were just doing the stuff that came on the QLib disk.
I have not done much with QLiberator in the last few days, But I have identified the format of the 'External Proc/Functions' data blocks (at least the parts I think I need to know). And added support in the decompiler to list the Procedures and Functions that need their names changed and a 'REMark $$external' added before them.
I still want to try to get QLIB to compile itself to an _OBJ file the same size as the original, and do a byte by byte comparison.