Re: Linux-2.4.10

[Erik Andersen]

On Mon Sep 24, 2001 at 07:48:47AM +0100, David Woodhouse wrote:
> 
> [e-mail] said:
> > Is jffs2 still showing the
> >     Child dir "." (ino #1) of dir ino #1 appears to be a hard link
> > problem?  I saw you patched mkfs.jffs2 after my changes -- do you
> > still need me to hunt down that bug I added? 
> 
> Yes please. The patch I committed just made it happier with a relative (or
> no) root directory - it was changing into the specified directory and then
> still prepending its name to every path. I assume it's still emitting a
> dirent for '.' in the root directory as it was before. The JFFS2 kernel code
> doesn't like that very much.

This seems to fix it, but I'm not certain this is correct?
Should / on jffs2 have neither inode nor dirent added?

--- mkfs.jffs2.c	2001/09/17 13:43:32	1.16
+++ mkfs.jffs2.c	2001/09/24 17:37:40
@@ -924,10 +924,11 @@
 		name = tmp_dir->name;
 		sb = tmp_dir->sb;
 		if (!tmp_dir->parent) {
+			/* Cope with the root directory */
 			ino = highest_ino++;
-			debug_msg("writing '/' ino=%lu", (unsigned long) ino);
-			output_pipe(ino, &sb);
-			write_dirent(ino, version++, ino, timestamp, DT_DIR, name);
+			debug_msg("writing '%s' ino=%lu", name, (unsigned long) ino);
+			//output_pipe(ino, &sb);
+			//write_dirent(ino, version++, ino, timestamp, DT_DIR, name);
 			tmp_dir = tmp_dir->next;
 			continue;
 		}

 -Erik

--
Erik B. Andersen   email:  [e-mail], formerly of Lineo
--This message was written using 73% post-consumer electrons--