[pkg-bacula-devel] git push for 9.0.5 ready

[Carsten Leonhardt]

Hi,

>> The only way I can think of how to address that problem cleanly is to
>> re-introduce bacula-director-common as a transitional package, but that
>> would only really help as a stable update.
>
> I can't comment on that, this is more your forté than mine. Does this
> mean this should also be added for Stretch via a point-release, as
> updates from Jessie to Stretch are also affected?

yes, in fact, adding it to unstable now is just "hoping for the best".

>> I'll probably upload the current state later, testing migration is
>> blocked by mariadb anyway...
>
> We should fix the "CacheRetention BIGINT UNSIGNED DEFAULT" confusion for
> mysql before that. I did the same for the sqlite3 case during merging,
> as you noticed and from looking at the code, I think "UNSIGNED" is the
> correct value here. So fixing update_mysql_tables.in to also use
> UNSIGNED would be correct, AFAICS.

My intuition says that the retention period will probably never be so
high that the highest bit will be needed.

> Postgres does not have "bigint UNSIGNED" because Postgres plainly does
> not have any unsigned data types, you have to "emulate" such data types
> by defining a domain with an additional constraint.

Good to know, so only mysql is left to be fixed.

So I'll wait until tomorrow before uploading, maybe Kern will answer to
my question on bacula-devel.

> I remember faintly there was a problem with MariaDB 10.2 and Bacula.
> Something about it being stricter for some Date or NULL values, can't
> remember fully right now.

I'm quite sure that Kern fixed that.

Carsten