Re: kadmin: failing dump/load
Patrik Lundin <[email protected]> Mon, 27 Nov 2017 13:34:25 +0100
| Newsgroups | gmane.comp.encryption.kerberos.heimdal.general |
|---|---|
| Message-ID | <[email protected]> |
On 2017-11-07 10:55:08, Patrik Lundin wrote: > On 2017-11-06 13:08:25, Viktor Dukhovni wrote: > > > > See: > > > > https://github.com/heimdal/heimdal/commit/d2130e3312089a3142e89b316d2900fa21855726 > > > > Interesting! I notice that patch does not apply to 7.4.0 as-is, is there > a new release planned soon or am I better off spending time getting it > to apply to the current release? > Just a quick follow-up regarding this: I went ahead and investigated how tricky it would be to apply that fix on 7.4.0, and it turned out to be fairly straight forward. The main problem was that hdb_set_sync() was introduced prior to the fix, but including it ended up with this set of patches: === e3cc7dfb94826aa709a617fb8e59ab07db584f92: Set def page size for HDB SQLite3 backend to 8KB d6979fcc40e00fd6ddd393c724cd728349b86de8: Allow LDAP modifications for entry creation 5bcbe2125b18160f6ad348b15f8036ffedc15770: Add hdb_set_sync() method 7d5f8bb051ca84592d1196bf5d5522da5a50f9d6: Disable sync during kadmin load d2130e3312089a3142e89b316d2900fa21855726: Handle long lines in dump files === Just looking around it seems it might be suitable to also add in 305dc816525f461f9bfe640d87f671f53f0e0fc6 (Disable sync during iprop receive_everything()), but I am trying to keep the customization to a minimum. Applying them in the listed order meant there was no patch fuzz, and I just now managed to test a load of a large database dump that was previously failing. It now finished without complaints, and a follow-up dump produced a file with a sha256 hash matching the original backup file. Thanks a bunch for fixing the problem. -- Patrik Lundin