Re: [Monotone-devel] Re: Monotone Bug with OpenEmbedded Database

[Otto Bommer]

I actually downgraded to monotone 0.30 for the Open Embedded DB which
works as this DB is maintained with OE 0.30. So I guess the DB is OK,
but it is actually as Koen said:
> Note for the monotone developers: can you *please* teach monotone the
difference between
> "too new" and "too old" when issueing db migrate messages? I've now
had to explain to
> multiple users that they shouldn't follow the advice monotone gives,
but the exact reverse
> (update mtn, not the db).

I put a remark on the OE homepage to use the same mtn version as the DB
and not use db migrate.
So i guess the issue is closed as far as monotone-devel is concerned.
Sorry for the noise.

Regards

Zack Weinberg wrote:
>> Note for the monotone developers: can you *please* teach monotone the
>> difference between
>> "too new" and "too old" when issueing db migrate messages? I've now
>> had to explain to
>> multiple users that they shouldn't follow the advice monotone gives,
>> but the exact reverse
>> (update mtn, not the db).
>
> I agree with this and am looking into how to do it now, but it's
> off-target in this case.  The "db migrate" command fails if it doesn't
> recognize the starting-point schema, monotone 0.25 would have given an
> unrecognized command error on "db regenerate_caches", and the original
> poster  reports --full-version output from 0.32.
>
> On the other hand, the failure mode (CRC32 error during decompression)
> makes me think that maybe the problem is data corruption in the
> database provided by the OE people, not a bug in monotone.  (We have,
> in general, a problem with misreporting exceptions thrown by 3rd-party
> libraries as monotone bugs.)
>
> Otto: since you seem to have mtn 0.32 already, perhaps you should just
> grab the this-is-for-mtn-0.32 snapshot from the OE site?
>
> zw