Bug#940521: buster-pu: package mariadb-10.3 10.3.18-0+deb10u1

2019-09-17 Thread Otto Kekäläinen
> Some poking around on Github suggests that the fix is
> https://github.com/MariaDB/server/commit/47f8a18fec604983e47fdf7c822d94b26d85cade
>
> If this is actively breaking systems, couldn't we simply apply that
> patch to the current versions in stretch and buster, as a least-change
> option?

We could in theory extend the gitlab-ci.yml pipeline to include a test
that sets up a replication master and a slave, and runs some updates
and replicates them. With that pipeline we could validate that earlier
versions worked and with this patch it works again. And such an
addition to the CI pipeline would also discover if replication ever
breaks again. This is however too much work for me right now, so
instead I would just wait until upstream releases and trust that
upstream knows what they are doing in terms of making the fix
correctly without breaking anything more somewhere else.



Bug#940521: buster-pu: package mariadb-10.3 10.3.18-0+deb10u1

2019-09-17 Thread Adam D. Barratt

On 2019-09-17 14:57, Otto Kekäläinen wrote:
[...]

Please go ahead.


OK, uploading today.


How are we looking for the 10.1 update for stretch?


Upstream has not yet released anything:
https://mariadb.com/kb/en/library/release-notes-mariadb-101-series/

I will get a release announcement email when upstream releases and
prepare the Debian upload in a couple days after that.


Some poking around on Github suggests that the fix is 
https://github.com/MariaDB/server/commit/47f8a18fec604983e47fdf7c822d94b26d85cade


If this is actively breaking systems, couldn't we simply apply that 
patch to the current versions in stretch and buster, as a least-change 
option?


Regards,

Adam



Bug#940521: buster-pu: package mariadb-10.3 10.3.18-0+deb10u1

2019-09-17 Thread Otto Kekäläinen
> [For reference, X-Debbugs-Cc is generally better than your mail client's
> CC when submitting bugs, as it means the recipient gets the mail from
> the BTS with the bug number included]

[rant] I'll try to remember, but this UI that is based on me
remembering all the option names and syntaxes (or having to look them
up from a separate page) usually leads to me using just the most
simple control words. Maybe someday Debian will have a UI that will
take away the requirement to remember so much.. [/rant]

> Please go ahead.

OK, uploading today.

> How are we looking for the 10.1 update for stretch?

Upstream has not yet released anything:
https://mariadb.com/kb/en/library/release-notes-mariadb-101-series/

I will get a release announcement email when upstream releases and
prepare the Debian upload in a couple days after that.

- Otto



Bug#940521: buster-pu: package mariadb-10.3 10.3.18-0+deb10u1

2019-09-17 Thread Adam D. Barratt

Control: severity -1 normal
Control: tags -1 -moreinfo +confirmed

[For reference, X-Debbugs-Cc is generally better than your mail client's 
CC when submitting bugs, as it means the recipient gets the mail from 
the BTS with the bug number included]


On 2019-09-16 20:08, Otto Kekäläinen wrote:

MariaDB 10.3.17 included a severe regression that broke replication in
setups where it is used. See details in #939819. As per suggested
there by Adam, we could upload a stable update to fix this.

There was no bug in Debian packaging. Changes only include upstream 
changes.


Changelog:

mariadb-10.3 (1:10.3.18-0+deb10u1) buster; urgency=high

  * New upstream version 10.3.18. Fixes regression introduced in 
10.3.17

(MDEV-20247: Replication hangs with "preparing" and never starts)
(Closes: #939819)


I do wish it were possible to get more targetted fixes for things like 
this. :-|


Please go ahead.

How are we looking for the 10.1 update for stretch?

Regard,

Adam