Re: [Bacula-users] Bacula 1.38.11 on OpenBSD 4.0

[Kern S. <ke...@si...>]

On Sunday 24 December 2006 23:12, Jeff Richards wrote:
> I think I saw something about this on the mailing list archive.  Don't know 
if anyone else has commented on this so here goes.
> 
> I was able to get Bacula compiled and working on OpenBSD 4.0 by 
copying /usr/include/sys/mtio.h from OBSD 3.9 to a temporary location and 
changing the reference to mtio.h in stored to point to the old copy.  When I 
did a diff of the two mtio header files it looks like the major changes 
between 3.9 and 4.0 was the removal of some structure elements which are "not 
yet implemented".  Here's the diff between the 3.9 and 4.0 header.
> 
> 1c1
> < /*    $OpenBSD: mtio.h,v 1.6 2003/06/02 23:28:21 millert Exp $        */
> ---
> > /*    $OpenBSD: mtio.h,v 1.7 2006/06/01 05:42:01 krw Exp $    */
> 75,78d74
> < /* the following two are not yet implemented */
> <       daddr_t mt_fileno;      /* file number of current position */
> <       daddr_t mt_blkno;       /* block number of current position */
> < /* end not yet implemented */
> 81,82c77,78
> <       daddr_t mt_mblksiz[4];  /* block size for different modes */
> <       daddr_t mt_mdensity[4]; /* density codes for different modes */
> ---
> >       daddr_t mt_mblksiz;     /* default block size */
> >       daddr_t mt_mdensity;    /* default density code */
> 
> My Bacula server does not go directly to tape but spools to disk.  I later 
off-load my disk images to tape.

Thanks for the information.  I'm pleased to hear that the mtio.h header is the 
only thing blocking building the Bacula SD on OpenBSD.

After some discussion of this matter, I believe that the OpenBSD guys are 
planning to put the two fields important for Bacula (mt_fileno, and mt_blkno) 
back into the header file, which will make compiling future versions of 
Bacula easier.  At that point, with the proper Device definition (UseMTIOCGET 
= no and some others) Bacula may even be able to write to tape.  

Best regards,

Kern