| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 2909.1 | | WJG::GUINEAU | Impossible Concentration | Tue Sep 12 1989 12:15 | 5 |
| this months transactor has an example on using LoadSeg() et al.
buried in the "Creating an Amiga Device" article.
John
|
| 2909.2 | But where is it? | NBOSWS::FRIES | | Tue Sep 12 1989 12:26 | 6 |
| I'm sorry of beeing an inexperienced guy.
So, where can I get 'this months transactor'?
Thanks
Gerald
|
| 2909.3 | | WJG::GUINEAU | Impossible Concentration | Tue Sep 12 1989 13:25 | 10 |
| Sorry.
It's a magazine called "Transactor For Amiga". Comes from somewhere in Canada.
Should be on news stands that carry Amiga mags.
If you wait till tonight, I'll type the short example in for you.
John
|
| 2909.4 | some docs to help | WJG::GUINEAU | Impossible Concentration | Wed Sep 13 1989 00:01 | 47 |
|
Here's a bit of the example. It was geared towards loading and starting a
device driver but here are the relevant parts for LoadSeg()
#include <exec/types.h>
#include <exec/libraries.h>
#include <exec/resident.h>
#include <exec/devices.h>
#include <libraries/dos.h>
#include <libraries/dosextens.h>
ULONG SegList, /* BCPL seglist pointer */
codeloc; /* pointer to code */
main()
{
(device stuff ommited)
SegList = (ULONG)LoadSeg"mydev.device"); /* load code into memory */
if(!SegList) {
printf("Can't load code!\n");
exit(0);
};
/* make SegList BCPL pointer into a real pointer */
codeloc = SegList * 4;
(rest of device stuf ommited)
}
Also, I'm probably going to get into trouble for this (CB, what's the story
here?) but here are some docs from the 1.3 AutoDocs. WJG::AMIGA:DOCS.ZOO
will hang around till someone tells me to cut the sh*t :-)
This ZOO has docs for AddTask(), RemTask(), CreateProc(), LoadSeg(), UnloadSeg()
John
|
| 2909.5 | CopyWrong ;-) | FRAMBO::BALZER | Christian Balzer DTN:785-1029 | Wed Sep 13 1989 03:36 | 17 |
| Re: .4
You (and other EasyNet archive maintainers) could be in deep trouble
indeed, if someone ever found out.
Now I'm not going to tell them, but...
Gerald, since your node name indicates that you're in Germany, go
on and gimme a call, will ya? :-)
A nice way to get all the docs and a bit of support is by becoming
a registered developer, which is relativly inexpensive...
BTW, you might be in for similar trouble with Transactor for
"duplicating" their copyrighted material here, John. ;-)
Regards,
<CB>
|
| 2909.6 | huh? Big Trouble? | WJG::GUINEAU | Impossible Concentration | Wed Sep 13 1989 08:09 | 4 |
|
Gone like the wind [docs.zoo]. Sorry!
John
|
| 2909.7 | | NBOSWS::FRIES | | Wed Sep 13 1989 11:02 | 9 |
| Thanks for your example, John.
But how to start that application (not a device or handler)
loaded by LoadSeg()? CreateProc() seems not to work (see 2909.0).
CreateProc() needs the segment list as a BCPL pointer, which LoadSeg()
supplies, I think. So segment_pointer<<=2L is not necessary (?)
Gerald
|
| 2909.8 | | WJG::GUINEAU | Impossible Concentration | Wed Sep 13 1989 11:50 | 20 |
|
If your example of CreateProc() was just as you used it,then that's the problem.
(i.e. you had CreateProc(seglist) - 1 arg)
CreateProc() takes several args, the seglist being one (and not the first).
I deleted the DOCS.ZOO file out of <CB> instilled fear :-) But I could
get the synopsis of CreateProc() for you.
Also, that example used an offset of 8 bytes passed what the SegList pointer
was as the actual start of the code.
SegList = LoadSeg("program");
codeloc = SegList * 4;
codestart = codeloc + 8;
John
|
| 2909.9 | Complete Example | NBOSWS::FRIES | | Thu Sep 14 1989 03:46 | 34 |
| Sorry.
Here is the complete example (which does not work).
struct Segment *seg,*LoadSeg();
main()
{
printf("main1\n");
NewTask();
Delay(300L); /* why not ? */
printf("main2\n");
/* UnLoadSeg(seg); */
exit(0);
}
NewTask()
{
seg = LoadSeg("my_program"); /* my_program is only a
* printf("Hello World\n");
* or such.
*/
CreateProc("MYPROG",0L,seg,4000L);
/* Now there is a task called MYPROG, but the application
* does NOT start.
*/
}
What's the mistake here (Or isn't it so easy) ?
Thanks
Gerald
|
| 2909.10 | | WJG::GUINEAU | Impossible Concentration | Thu Sep 14 1989 09:04 | 4 |
| I'll try this out tonight after school and see if I can figure it out.
John
|
| 2909.11 | | MILTSS::MILOS | Milos Roberto TSC Milan | Thu Sep 14 1989 15:23 | 11 |
| re: .9
I think your problem could be what you expected from your
"application".
I suppose that a printf in the task "MYPROG" will not write on *YOUR*
cli window.
What about do a "fprintf" into some log file to see if the code is
ever executed?
Good coding....8-)
Roberto.
|
| 2909.12 | the answer | WJG::GUINEAU | Impossible Concentration | Sat Sep 16 1989 14:00 | 59 |
|
> Why does th following code not work?
> seg=LoadSeg("file");
> CreateProc(seg);
> ...
> UnLoadSeg(seg);
Well, I finally figured this one out.
Basically what was happening is that the program you LoadSeg()'ed
(my_program) is compiled with the standard c startup module (c.o and _main)
by default.
The "normal" flow of a compiled program is like this:
<run program>
start is called which is in c.o. start does some stuff for
getting a vector area set up, opens dos.library for you and
deals with ARGC and ARGV for CLI or WB startup and then calls
_main (UNDERSCORE main)
_main is called (which is in umain.c and is added in to your program
when you compile it by lattice C). _main does some more
ARGC,ARGV stuff, again depending on CLI or WB startup. It deals
with stdin and stdout (making a small window if run from WB)
and finally calls your main().
main does what your program was wtitten to do!
So you LoadSeg my_program. CreateProc calls it, which causes it to execute
the standard startup code - except this program (my_program) was not run in
the "normal" way (i.e. not run from CLI or WB) so, because of reasons beyond
this explination, the startup code thinks it was started from WorkBench - and
hangs in a WaitPort() call waiting for the startup message from workbench.
Your main never gets called!!!
The solution? HACK c.a and umain.c to death. For Lattice these are found
in LC:/Source/c.a, umain.c.
What I did was strip c.a down so all it does is set up the vector stuff,
open dos.library for you and call your main(). _main() (in umain.c) NEVER
gets called!
The next reply is the c.a I hacked up and the reply after that is
my_program.c that works.
REMEMBER!!! my_program executes in a very limited environment! No stdin, no
stdout, no NADA! All you have is dos.library so file I/O is ok. Intuition
stuff should also work fine so you can open windows and all. Also, ARGC and
ARGV are non existant.
Have Fun!
John
|
| 2909.13 | HACKED c.a startup code | WJG::GUINEAU | Impossible Concentration | Sat Sep 16 1989 14:01 | 171 |
| *
* Johns HACKED C initial startup procedure under AmigaDOS
*
* Use the following command line to make c.o
* asm -u -iINCLUDE: c.a
*
INCLUDE "exec/types.i"
INCLUDE "exec/alerts.i"
INCLUDE "exec/nodes.i"
INCLUDE "exec/lists.i"
INCLUDE "exec/ports.i"
INCLUDE "exec/libraries.i"
INCLUDE "exec/tasks.i"
INCLUDE "exec/memory.i"
INCLUDE "exec/execbase.i"
INCLUDE "libraries/dos.i"
INCLUDE "libraries/dosextens.i"
INCLUDE "workbench/startup.i"
INCLUDE "exec/funcdef.i"
INCLUDE "exec/exec_lib.i"
INCLUDE "libraries/dos_lib.i"
MEMFLAGS EQU MEMF_CLEAR+MEMF_PUBLIC
AbsExecBase EQU 4
; some usefull macros:
callsys macro
CALLLIB _LVO\1
endm
xdef XCEXIT * exit(code) is standard way to leave C.
xdef @XCEXIT
xref LinkerDB * linker defined base value
xref _BSSBAS * linker defined base of BSS
xref _BSSLEN * linker defined length of BSS
IFD RESIDENT
xref RESLEN
xref RESBASE
xref NEWDATAL
ENDC
* library references
section text,code
xref main * go strait to main(), do not pass go...
xref _main * Name of C program to start with.
xref MemCleanup * Free all allocated memory
xref __fpinit * initialize floating point
xref __fpterm * terminate floating point
start:
move.l a0,a2 * save command pointer
move.l d0,d2 * and command length
lea LinkerDB,a4 * load base register
lea _BSSBAS,a3 * get base of BSS
moveq #0,d1
move.l #_BSSLEN,d0 * get length of BSS in longwords
bra.s clr_lp * and clear for length given
clr_bss move.l d1,(a3)+
clr_lp dbf d0,clr_bss
move.l AbsExecBase.W,a6
move.l a6,SysBase(A4)
move.l a7,_StackPtr(A4) * Save stack ptr
clr.l WBenchMsg(A4)
*------ get the address of our task
move.l ThisTask(a6),A3
*----- clear any pending signals
moveq #0,d0
move.l #$00003000,d1
callsys SetSignal
move.l a7,D0 * get top of stack
sub.l 4(a7),D0 * compute bottom
add.l #128,D0 * allow for parms overflow
move.l D0,_base(A4) * save for stack checking
*------ attempt to open DOS library:
bsr.w openDOS
jsr __fpinit(PC) * Initialize floating point
jsr main(PC) * call C entrypoint
moveq.l #0,d0 * set successful status
bra.s exit2
*
XCEXIT:
move.l 4(SP),d0 * extract return code
@XCEXIT:
exit2:
move.l d0,-(a7)
move.l _ONEXIT(A4),d0 * exit trap function?
beq.s exit3
move.l d0,a0
jsr (a0)
exit3 jsr MemCleanup(PC) * cleanup leftover memory alloc.
move.l AbsExecBase.W,a6
move.l DOSBase(A4),a1
callsys CloseLibrary * close Dos library
jsr __fpterm(PC) * clean up any floating point
done_1c:
*------ this rts sends us back to DOS:
exitToDOS:
MOVE.L (A7)+,D0
movea.l _StackPtr(a4),SP * restore stack ptr
rts
*-----------------------------------------------------------------------
noDOS:
moveq.l #100,d0
bra exit2
*-----------------------------------------------------------------------
* Open the DOS library:
openDOS
lea DOSName(PC),A1
moveq.l #0,D0
callsys OpenLibrary
move.l D0,DOSBase(A4)
beq noDOS
rts
DOSName dc.b 'dos.library',0
section __MERGED,BSS
*
xref DOSBase
xdef NULL,SysBase,WBenchMsg
xdef curdir,_mbase,_mnext,_msize,_tsize
xdef _oserr,_OSERR,_FPERR,_SIGFPE,_ONERR,_ONEXIT,_ONBREAK
xdef _SIGINT
xdef _ProgramName,_StackPtr,_base
*
NULL ds.b 4 *
_base ds.b 4 * base of stack
_mbase ds.b 4 * base of memory pool
_mnext ds.b 4 * next available memory location
_msize ds.b 4 * size of memory pool
_tsize ds.b 4 * total size?
_oserr equ *
_OSERR ds.b 4
_FPERR ds.b 4
_SIGFPE ds.b 4
_SIGINT ds.b 4
_ONERR ds.b 4
_ONEXIT ds.b 4
_ONBREAK ds.b 4
curdir ds.b 4
SysBase ds.b 4
WBenchMsg ds.b 4
_StackPtr ds.b 4
stdin ds.b 4
_ProgramName ds.b 4
END
|
| 2909.14 | my_program.c | WJG::GUINEAU | Impossible Concentration | Sat Sep 16 1989 14:11 | 38 |
|
/*
* my_program.c - intended to be LoadSeg'ed and CreateProc'ed from
* another program.
*
* lc my_program
* blink with my_program.lnk
*
* my_program.lnk:
*
* FROM "c.o"+"my_program.o"
* TO "my_program"
* LIB LIB:lc.lib LIB:amiga.lib
* BATCH
*
* NOTICE the FROM "c.o" which is the compiled version of the previous reply.
* Normally this is LIB:c.o which is Lattice's standard startup code.
*
*/
#include <stdio.h>
main()
{
FILE *fp;
fp = fopen("ram:test","w");
if(!fp)
exit(10);
fprintf(fp,"Hi");
fclose(fp);
exit(0);
}
|
| 2909.15 | GREAT !!! | NBOSWS::FRIES | | Mon Sep 18 1989 03:36 | 6 |
| GREAT !!!!
Thank you VERY much.
I hope Aztec will work, too.
Gerald
|
| 2909.16 | | WJG::GUINEAU | Impossible Concentration | Mon Sep 18 1989 08:57 | 10 |
| Your welcome. I learned alot doing it!
I would think Aztec should work fine. You might need to change some formatting
for c.a, but other than that, who knows!
BTW - There may even be a switch to say "toss the standard startup code
and use mine" or something like that. Check it out...
John
|
| 2909.17 | To simulate a startup-message ? | NBOSWS::FRIES | | Fri Sep 22 1989 03:31 | 14 |
| re .11:
No, 'fprintf(fp,"Hello World\n")' doesn't work, too.
re .12/13:
Again: Great work, John.
But in this way, perhaps it's not possible to start 'commercial'
applications like compilers and such.
Isn't there a way to 'simulate' a startup-message from Workbench
after a call to 'CreateProc()' and all works fine ?
So, how to call the PutMsg() ?
Gerald
|
| 2909.18 | Browser, WBRun... | FRAMBO::BALZER | Christian Balzer DTN:785-1029 | Fri Sep 22 1989 03:46 | 16 |
| Re: .17
There is a program including sources called WBRun around, and something
similar (also including sources) from Peter da Silva, which he uses
to start applications from his Browser (Version 1.6). This source
_should_ be all you need and it's on a Fish disk somewhere. If you
can't locate it, I'll be able to tell you where to get it on Monday.
Regards,
<CB>
P.S.
My vice-president (Heiko Rath) is still waiting for your call. ;-)
|
| 2909.19 | | WJG::GUINEAU | Impossible Concentration | Fri Sep 22 1989 09:06 | 5 |
|
WJG::AMIGA: has WBRUN (source) and BROWSER16 (nosource)
john
|
| 2909.20 | No I/O from task | ELMST::VERMA | Virendra, MRO4-3/H10, DTN 297-5913 | Tue Oct 03 1989 10:08 | 6 |
| I think you can't call Amiga DOS functions (such as I/O) from a task. This
is because I/O functions require more data structures that are not included
in the task structure. If your task needs to call Amiga DOS functions, a
process should be used rather than task.
Hope that helps.
|