> maximum entropy <entropy@tappedin.com> writes:
> > I submitted a patch for this problem in PR port-i386/13399 last July.

I'd just like to report that I applied Hr entropy's patch to my source
tree, and successfully built a kernel that works spiffy on the problem
machines I have.


However, the rest of the world build didn't go so well, with deltas from
about 18h on 06.feb being the latest to go into my tree, and I ended up
with a corrupted src/usr.sbin/makefs/ffs/mkfs.c,v RCS file.  This file
was obtained by cvsup from norway, so the problem may be on that mirror
if not on the site it mirrors, and the file was updated today in my local
repository.  The ftp and cvsweb mirrors I checked have unblemished copies
of this particular file...

    75	@/*	$NetBSD: mkfs.c,v 1.9 2002/01/31 22:44:04 tv Exp $	*/
    76	/* From NetBSD: mkfs.c,v 1.59 2001/12/31 07:07:58 lukem Exp $	*/
    77	
[...]
   187	union {
   188		struct fs fs;
   189		char pad[SBSIZE];
   190	} fsun;
   191	#define	sblock	fsun.fs
   192	
   193	union {
   194		struct cg cg;
   195		char pad[MAXBSIZE];
   196	} cgun;
   197	#define	acg	cgun.cg
   198	
   199	struct dinode zino[MAXBSIZE / DINODE_SIZE];
[snippity]
   216	
   217	union {
   218		struct fs fs;
   219		char pad[SBSIZE];
   220	} fsun;
   221	#define	sblock	fsun.fs
   222	
   223	union {
   224		struct cg cg;
   225		char pad[MAXBSIZE];
   226	} cgun;
   227	#define	acg	cgun.cg
   228	
   229	struct dinode zino[MAXBSIZE / DINODE_SIZE];
   230	
[more snipping]
   246	static	int	cpgflg;		/* cylinders/cylinder group flag was given */
   247	static	int	minfree;	/* free space threshold */
   248	static	int	opt;		/* optimization preference (space or time) */
   249	static	int	density;	/* number of bytes per inode */
   250	static	int	maxcontig;	/* max contiguous blocks to allocate */
   251	statectors/track including spares */
   252	static	int	secpercyl;	/* sectors per cylinder */
   253	static	int	sectorsize;	/* bytes/sector */
   254	static	int	rpm;		/* revolutions/minute of drive */
   255	static	int	interleave;	/* hardware sector interleave */
   256	static	int	trackskew;	/* sector 0 skew, per track */
   257	static	int	fsize;		/* fragment size */
   258	static	int	bsize;		/* block size */
   259	static	int	cpg;		/* cylinders/cylinder group */
   260	static	int	cpgflg;		/* cylinders/cylinder group flag was given */
   261	static	int	minfree;	/* free space threshold */
   262	static	int	opt;		/* optimization preference (space or time) */
   263	static	int	density;	/* number of bytes per inode */
   264	static	int	maxcontig;	/* max contiguous blocks to allocate */
   265	static	int	rotdelay;	/* rotational delay between blocks */
[you get the drift]

This might be one of the files that I saw indicating a `checksum mismatch --
transferring entire file' but I didn't think to log that cvsup session, and
if so, then I have a bunch more such in my tree.



> Sometimes this happens -- the trick is to bitch about it when it does
> and not to let it languish. We don't have full time staff in the
> NetBSD project so some of the PRs languish if no one champions them.

Stand back -- THIS ONE'S MINE