Execute() do not works under 1.3

[TCH]

I am trying to run an external program like this

Execute("programname", 0, 0);

and under 2.0+, it works perfectly, but under 1.3 it returns immediately without running the program. I've tried to open

dos.library

before it

struct Library *dos = OpenLibrary("dos.library", 33);
if (dos == NULL)
{
	fprintf(stderr, "OpenLibrary() failed.\n");
	return 1
}

but nothing has changed. (Under 2.0+, it worked without this.)

What is the difference? I assume i have to open a library under 1.3, but which?

[ross]

Do you have 'Run' in C:?

Under KS1.x is mandatory.

[TCH]

No i did not have it, i did not know, that under 1.3

Execute()

relies on an external program. Thanks for pointing out: now it works under 1.3.

[Galahad/FLT]

Originally Posted by TCH

No i did not have it, i did not know, that under 1.3

Execute()

relies on an external program. Thanks for pointing out: now it works under 1.3.

"Run" as well as a list of other programs were integrated into Kickstart 2.0+

[TCH]

I've noticed that WB3.1 did not have a

Run

command, but i did not know, that

Execute()

calls

Run

and not the other way.

[Thomas Richter]

Originally Posted by TCH

I've noticed that WB3.1 did not have a

Run

command, but i did not know, that

Execute()

calls

Run

and not the other way.

Actually, that is only the case for (the BCPL-implemented) dos.library in v34 and below. The logic changed in v36 and the C/asm version of the dos.library. There, Execute() calls into System(), and so does C:Run. Execute() does not depend on C:Run anymore.

[TCH]

I see, thanks.

[alkis]

I have a routine, which relies on

*****   SYSTEM0 execute command from disk. From Sozobon C distribution.
* Copyright © 1988 by Ralph Babel.
*
* This piece of code may be used as part of any product as
* long as the source code for the complete program can be
* obtained free of charge (except for a small copying fee)
* and this copyright notice is left unchanged.

it works on everything (1.2-3.1 tried at least), and doesn't need Run.

Exposes this mysystem(char *cmd)
(define NDEBUG to lose all the debugging msgs send to serial)

Has a stupid bug I can't seem to iron out. If command is resident then it crashes when my code tries

    /*
    Forbid();
    if (list->count > 0)
      list->count--;
    Permit();
    */

which is why I have it commented out. Other than that, it works
Hope it helps.

[redblade]

Love those people who share their source

[TCH]

@alkis:

Thank you for sharing your source. Although i have already released my program, i will study your code to get a better view on the mechanism of running something on the Amiga.

Furthermore, i would like to re-share it on my website; could you please tell me the last year you have modified the files? Are the file dates correct in the zip archive? (2009 for the C and 2014 for the ASM file.) I need this, because i display a release date for downloads.

[alkis]

You are welcome. Both files have modification date in 2019.

unzip -v system.zip 
Archive:  system.zip
 Length   Method    Size  Cmpr    Date    Time   CRC-32   Name
--------  ------  ------- ---- ---------- ----- --------  ----
    8713  Defl:N     3166  64% 2019-11-09 16:38 aa5e684d  mysystem.c
   11274  Defl:N     4162  63% 2019-04-14 19:49 8dca550b  system0.asm
--------          -------  ---                            -------
   19987             7328  63%                            2 files

the two timestamps there agree with my filesystem's.

[TCH]

Weird.

unzip

gave these dates for me too, but Ark not:



It seems, that it interprets the date in a wrong order. But that is not so surprising, as Ark sucks in a lot of aspects.

I've uploaded your sources to my website, thanks again.

[matburton]

In case it's interesting I also got caught out by Execute needing Run

What a PITA!

[alkis]

totally off-topic, but Ark 17.12.3 displays the correct timestamps (kubuntu 18.04.3)

[TCH]

@matburton:

In the thread you were told, that you only should edit the list between

Forbit()

and

Permit()

, but alkis was exactly tried to do that and his code has crashed. Did you experience the same? (I do not understand why Leffman called

BSTR

horror: a leading byte what holds the length for the text-chunk makes much more sense than terminating it with a

NUL

; ANSI strings have indefinite lengths by definition, which i always considered as crazy. That's why

strlen()

in C is "expensive", it has to iterate the entire data block, just to know the size. And also, it means, that you cannot use

NUL

characters in ANSI strings... But this is off here.)

@alkis:

This is Ark 2.6.4 from TDE. I guess they do not paid too much attention to Ark, so this is not fixed.