PedroM 0.82 - Page 5

While looking at Link.asm, I noticed two "clr" without any size specification (./76). I think this may be undesirable.

Thanks.
Fixed.

PpHd (./96) :
the RAM_THROW extension

Could we have official macros to use that ? I'll do it myself if you don't want to, but if there is nothing official, it could result in a mess with sources of different developpers.

Folco (./123) :

PpHd (./96) :
the RAM_THROW extension

Could we have official macros to use that ? I'll do it myself if you don't want to, but if there is nothing official, it could result in a mess with sources of different developpers.

There is no official assembly include for PedroM anyway
And don't be afraid; I am pretty sure you'll be the one and the only one who use this macro

Do you want official ones, for GAS and A68k, including the macros and this one :

.set	pedrom__stdin,		pedrom__0000
.set	pedrom__stdout,		pedrom__0000+4
.set	pedrom__stderr,		pedrom__0000+8

.set	pedrom__printf,		pedrom__0004
.set	pedrom__vcbprintf,	pedrom__0005
.set	pedrom__clrscr,		pedrom__0006
.set	pedrom__fclose,		pedrom__0007
.set	pedrom__freopen,	pedrom__0008
.set	pedrom__fopen,		pedrom__0009
.set	pedrom__fseek,		pedrom__000a
.set	pedrom__ftell,		pedrom__000b
.set	pedrom__feof,		pedrom__000c
.set	pedrom__fputc,		pedrom__000d
.set	pedrom__fputs,		pedrom__000e
.set	pedrom__fwrite,		pedrom__000f
.set	pedrom__fgetc,		pedrom__0010
.set	pedrom__fread,		pedrom__0011
.set	pedrom__fgets,		pedrom__0012
.set	pedrom__ungetc,		pedrom__0013
.set	pedrom__fflush,		pedrom__0014
.set	pedrom__clearerr,	pedrom__0015
.set	pedrom__ferror,		pedrom__0016
.set	pedrom__rewind,		pedrom__0017
.set	pedrom__fprtinf,	pedrom__0018
.set	pedrom__tmpnam,		pedrom__0019

.set	pedrom__qsort,		pedrom__001b

.set	pedrom__bsearch,	pedrom__001e
.set	pedrom__remove,		pedrom__001f
.set	pedrom__unlink,		pedrom__001f
.set	pedrom__rename,		pedrom__0020
.set	pedrom__atoi,		pedrom__0021
.set	pedrom__atol,		pedrom__0021

.set	pedrom__rand,		pedrom__0023
.set	pedrom__srand,		pedrom__0024
.set	pedrom__calloc,		pedrom__0025
.set	pedrom__realloc,	pedrom__0026
.set	pedrom__atof,		pedrom__0027
.set	pedrom__sttputchar,	pedrom__0028	| ???
.set	pedrom__perror,		pedrom__0029
.set	pedrom__getenv,		pedrom__002a
.set	pedrom__system,		pedrom__002b
.set	pedrom__setvbuf,	pedrom__002c
.set	pedrom__exit,		pedrom__002d
.set	pedrom__atexit,		pedrom__002e

?

And don't be afraid; I am pretty sure you'll be the one and the only one who use this macro

BTW, many thanks for having merged the code of this extension.

Folco (./125) :
?

Here is a simple, relative, and fast answer: non
Propose this patch to gcc4ti instead

Well.

Folco (./127) :
Well.

Where is your awaited joke?

No, all is right, I'm already documenting all pedrom@00xx for the one who could write the C header.

RC9 has been released (at the usual place). It fixes:
+ the problem of getting a file.
+ getting the list of files on V200.
It adds a new minor export (Side font).

I'm still waiting for tests with ti-connect.

PpHd (./130) :
It adds a new minor export (Side font).

Great! Will be availible via FontSetSys?

Wonderful !
Uther -> No, because a lot of functions should be rewriten. But you can use it, its ptr should be at .set pedrom__0000+28 (I'm right ?)

PpHd -> I'm still waiting a version of TiLP which would allow to upload log files from the calc without crashing... I have still tried yesterday...

wow Patrick vu l'heure du build t'es un lève-tôt

Uther (./131) :

PpHd (./130) :
It adds a new minor export (Side font).

Great! Will be availible via FontSetSys?

As Folco said, it is a lot more work to use it in all the Draw functions. It won't be done for 0.82. Maybe 0.83.

Folco (./132) :
PpHd -> I'm still waiting a version of TiLP which would allow to upload log files from the calc without crashing... I have still tried yesterday...

You shouldn't get this behavious with this version. PedroM shouln't crash anymore tilp.

Folco (./132) :
wow Patrick vu l'heure du build t'es un lève-tôt

Mais non, mais non.

Suggestion : The display of the current folder name is done in upper case, but the filesystem is from now on case-sensitive.
I think it would be a good idea to disp the current folder name in the real case.

Patch (not tested, I can't compile...) :

Graph.asm

ST_folder:
	move.l	4(a7),a0
ST_folder_reg
	jsr	ST_IsHelpQuit
-	; Cvt to Upper case
	lea	FOLDER_TEMP,a1
	moveq	#8-1,d1
\loop		move.b	(a0)+,d0
		beq.s	\fill
-		cmpi.b	#'a'-1,d0
-		bls.s	\NoCvt
-		cmpi.b	#'z',d0
-		bhi.s	\NoCvt
-			add.b	#'A'-'a',d0	; Cvt to Upper case
-\NoCvt		move.b	d0,(a1)+
+		move.b	d0,(a1)+
		dbf	d1,\loop
		bra.s	\final
\fill		move.b	#' ',(a1)+
		dbf	d1,\fill
\final	move.b	#' ',(a1)+
	clr.b	(a1)+

Someone has an idea about ./134 ?

Fix done in future RC10:
+ A legend is displayed during 'install tib' command
+ PedroM is cooler about wrong protocol format (ie. accept both len of 7 and 6 for 89 and 92+) with the 'install tib' command.
+ PedroM expected a 92+/89 RTS header for V200 and 89TI, whereas it should expect a different one. Fix this.
+ Wait 1s before rebooting at the end of the transfert to be sure the OK message is sent.

Now I can use 'install tib' command with tiemu + tilp (with the 2 fixes I have explained in the other thread) for 89 / 92+ / 89TI / V200.
I have even try sending a 92+ TIB to a 89 titanium, and it works: tiemu crashes when the reboot is done!

If the filesystem is case-sensitive, ./134 looks like a good idea to me.
Something odd, though, but I guess that's because the patch was generated manually: I doubt there are two move.b d0,(a1)+.
If the conversion is removed, maybe you can use dbxx, as we've devised in another topic ?


As part of testing your patches, I have just mistakenly transferred a V200 TIB to a 89T. The transfer ran to completion, but of course, the 89T didn't boot. TIEmu didn't crash, however.
Did you save the backtrace of the crash ?

No. But I don't run a cutting edge tiemu.

Lionel Debroux (./136) :
Something odd, though, but I guess that's because the patch was generated manually

You're right, I have only Win Explorer and the Notepad here.

Lionel Debroux (./136) :
I doubt there are two move.b d0,(a1)+.

So perhaps it's not obvious to read...

 [b]Folco (./133) :[/b]
-\NoCvt		move.b	d0,(a1)
+  		move.b	d0,(a1)+

The first line removes the useless label. (I had omitted the "+". Fixed, thanks).

./134

the "bra.s \final" could replaced with "addq.w #1,d1", and the "\last: move.b #' ',(a1)+"removed, no ? (-2 bytes... ok, c'est pas byzance, mais voilà...)

Vectors.asm, RAM_THROW extension:

		bcs.s	\NotAFunctionCall
		cmp.l	#kernel::EndSharedLinker,a1	; Compare the address. Is-it inside the PreOs kernel routine?
		bhi.s	\NotAFunctionCall
-			btst.l	#5,d1			; Called in supervisor mode ? 
+			btst.l	#13,d1			; Called in supervisor mode ? 
			beq.s	\UserMode		; No 
				move.w	d1,6*4(sp)		; Rewrite SR (need another return adress on the stack) 
				move.l	a1,6*4+2(sp)		; Set the ramcall ptr as the return adress of the handler 

I had writen '5' because I tested the bit in the upper byte of the SR inside the stack ... I didn't see that when you optimized this part of the routine. Sorry for not having catch that...

In my programs (man & eexec), this doesn't produce a bug, because the bit 5 of the SR is not used (so it's 0), and I always call the handler when in user mode.

Ooooooooooouuuuuuuuuuuuppppppppppppppppppssssssssssssssssssssss
You're right.

Folco (./134) :
Suggestion : The display of the current folder name is done in upper case, but the filesystem is from now on case-sensitive.
I think it would be a good idea to disp the current folder name in the real case.

Patch (not tested, I can't compile...) :

Graph.asm

ST_folder:
	move.l	4(a7),a0
ST_folder_reg
	jsr	ST_IsHelpQuit
-	; Cvt to Upper case
	lea	FOLDER_TEMP,a1
	moveq	#8-1,d1
\loop		move.b	(a0)+,d0
		beq.s	\fill
-		cmpi.b	#'a'-1,d0
-		bls.s	\NoCvt
-		cmpi.b	#'z',d0
-		bhi.s	\NoCvt
-			add.b	#'A'-'a',d0	; Cvt to Upper case
-\NoCvt		move.b	d0,(a1)+
+		move.b	d0,(a1)+
		dbf	d1,\loop
		bra.s	\final
\fill		move.b	#' ',(a1)+
		dbf	d1,\fill
\final	move.b	#' ',(a1)+
	clr.b	(a1)+

TIGCC doc said that it shall be converted to upper case... What to do?

displays the given folder name (converted
to uppercase letters)

It's just a consequence, no ? a constatation ? no more I think, just to bemore readable... I think it can't matter for a user program

I hadn't noticed what you say in ./142, but IMO, a slight behavioral difference between ST_folder on PedroM (whose FS is case-sensitive nowadays, according to ./134) and on AMS (whose FS only understands lowercase symbol names, as indicated by the SymFind documentation - AFAICT, no AMS program is using uppercase in symbol names anyway) is no problem.

The mere fact that the filesystem is case-sensitive is enough to break user programs. AMS programs are written to expect a case-insensitive file system. Linking programs expect it too to some extent, for example, the file names show up in the file names on the computer file system, so on case-insensitive PC file systems, the files FOO, FOo, FoO, Foo, fOO, fOo, foO and foo will all have conflicting names.

1. I have performed this modification in the display of the folder.
2. I have fixed the bug preventing you from sending case identical filename.
3. I have included the command DEL ==> Create folder now works with TILP, but not delete var since it checks for AMS 2.09, which PedroM is not.
4. Fix on V200: The HW_VERSION to send in the VER command is the real HW_VERSION (ie. 2) not (1) (But WHY!????)
5. Return the archived flag for directory listing.
6. Fix the returned version by the VER command: the version should be in decimal, not hexadecimal.

The V200 is weird because it has gateArray=hardwareRevision=2. For the other models, gateArray=HW_VERSION and hardwareRevision=HW_VERSION-1.

PedroM 0.82 RC10 has been released.
I hope it is the last RC before the official release. If there is no critical bug, 0.82 will be released next week I think.
(Due to new TILP, I was able to send this RC PedroM to my calculator ==> Success for the testing suite except the batteries -not tested).

(vivement vendredi )

, indeed.

BTW: I need help testing AMS transfers (to a calc running PedroM, and to a calc running AMS) with TILP.