On Tuesday 02 February 2010 12:04:03 Graham Keeling wrote: > On Tue, Feb 02, 2010 at 11:59:43AM +0100, Kern Sibbald wrote: > > On Tuesday 02 February 2010 10:53:07 Graham Keeling wrote: > > > On Mon, Feb 01, 2010 at 04:35:20PM +0100, Kern Sibbald wrote: > > > > On Monday 01 February 2010 14:57:47 Graham Keeling wrote: > > > > > Hello Kern, > > > > > > > > > > Can you tell me where in the code the change was that prevents the > > > > > storage daemon crashing on reading my corrupt backup-0009? > > > > > > > > The first part was: > > > > Patch: 3d6f161a8f76799cde5ebbb3f06de965e5ad9085 > > > > made on 18Jan10 at 13:31:00 > > > > > > > > The Verify part is: > > > > > > > > Patch: ef59bf981a1d828e0c06a1c7187f3136a5d6857f > > > > made on 28Jan10 at 14:41:35 > > > > > > Hello, > > > > > > Those are to do with removing the index checks. > > > > > > But I thought there was another change to do with preventing the > > > storage daemon from crashing with a corrupted volume? > > > > Sorry, I didn't understand. Preventing a crash was submitted in several > > patches around the same time -- it is patches that relate to > > src/lib/serial.c and serial.h. It was committed in several stages -- > > the first was incorrect; I did a little too much and that broke some > > things; the second patch corrected that -- there may even have been a > > third. Since it requires a bit more research than I did for the previous > > response, I'll let you find them, or try version 5.0.0 directly. > > Yes, I've got it now. I found the patches and matched them up with what is > now in 5.0.0, to make sure I am up to date. >
By the way, I have been quite busy lately, but I think it would be worth while taking a careful look at the damage that was done in your volume (the one you sent me). If I remember right, it was in two of the "session" records, which makes me suspicious that there might be a bug somewhere -- of course, on the other hand, those are the only records that Bacula copies as "strings" so if there were damage elsewhere (but not in a record header, it would never be noticed. If you have some time, after applying the patch, you might try dumping out all the session records and see if the kind of damage makes any sense or has any systematic consistency which could point to a bug. Regards, Kern ------------------------------------------------------------------------------ The Planet: dedicated and managed hosting, cloud storage, colocation Stay online with enterprise data centers and the best network in the business Choose flexible plans and management services without long-term contracts Personal 24x7 support from experience hosting pros just a phone call away. http://p.sf.net/sfu/theplanet-com _______________________________________________ Bacula-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/bacula-devel
