Re: Improved autofix patch

[mk@anuurn.do.open.de (Martin Koehling)]

In <01HW5TDMJD2S004NHI@CATHY.IJS.SI>,
Matija Grabnar (Matija.Grabnar@ijs.si) wrote:

>Following comments abot the previous patch I posted here, I am posting
>a new patch. This one is a context diff (thanks for the tip).

My `patch.ttp' somehow doesn't like this diff...

What happens is this:

  N:\mint-1.12.4> patch <autofix_patch.dif
  Hmm...  Looks like a new-style context diff to me...
  The text leading up to this was:
  --------------------------
  |*** /f/freemint.12/main.c      Wed Aug  9 19:29:18 1995
  |--- main.c     Sat Oct  7 00:09:26 1995
  --------------------------
  Patching file main.c using Plan A...
  patch: **** unexpected end of hunk at line 57
  N:\mint-1.12.4>

Did anybody else have problems with this diff?
Or is my `patch' broken (or too old)? (It worked fine with all 
diffs I threw at it in the past, though, as long as correct line
terminators were used...)
Where can I find a more current version?
And of `diff', too?
(I'm currently using a `diff' from 1991 and a `patch' 1992 - both
clearly not MiNT-aware...:-()

(BTW: IMHO, whenever possible, a unified diff should be created - it's
both shorter and more readable than a normal context diff...)

>I do think that the cd command in MINT.CNF should be obeyed, but I also think
>AUTOFIX _must_ ensure that programs following MiNT get executed - otherwise
>it is too confusing for the users.

You're right on both points; since my own setup is more TOS-like
than Unix-like, I didn't quite see the advantages of being able
to set a  different directory for the INIT process...

I hereby formally revoke my original patch suggestion! :-)

Your patch still has a few problems, though; in particular:

- the current directory during auto folder execution should be
  root, *not* \AUTO
- we probably shouldn't intermix calls to "internal" MiNT functions
  like d_getdrv() and GEMDOS functions like Dgetdrv() (actually,
  this only matters if we're in kernel mode, which is not the case
  at this point [I think]; still, it's cleaner to always use the
  "internal" functions...)
- use PATH_MAX instead of some constant (64) for the path buffer
  size...

I appended a "fixed" and more streamlined version of your patch to this
message.

Martin

--- ..\mk-src-1\main.c	Thu Oct 12 18:22:10 1995
+++ main.c	Thu Oct 12 20:44:20 1995
@@ -1668,6 +1668,8 @@
 	long r;
 	static char pathspec[32] = "\\AUTO\\";
 	short runthem = 0;	/* set to 1 after we find MINT.PRG */
+	char	curpath[PATH_MAX];
+ 	int	curdriv,bootdriv;

 /* if the AES is running, don't check AUTO */

@@ -1676,6 +1678,15 @@
 /* OK, now let's run through \\AUTO looking for
  * programs...
  */
+	d_getpath(curpath,0);
+	curdriv = d_getdrv();
+	/*
+	*	We are in Supervisor mode, so we can do this
+	*/
+	bootdriv=*((short*)0x446);
+	d_setdrv(bootdriv);
+	d_setpath("\\");
+
 	dta = (DTABUF *)f_getdta();
 	r = f_sfirst("\\AUTO\\*.PRG", 0);
 	while (r >= 0) {
@@ -1692,4 +1703,6 @@
 		}
 		r = f_snext();
 	}
+ 	d_setdrv(curdriv);
+ 	d_setpath(curpath);
 }

-- 
 Martin Koehling // NEW email address: mk@anuurn.ruhr.de