Re: openssl 1.1 fallout with mmysql, mysql succession planning?

[Greg Troxel <gdt%lexort.com@localhost>]

Jonathan Perkin <jperkin%joyent.com@localhost> writes:

> * On 2020-01-19 at 22:15 GMT, Greg Troxel wrote:
>
>> With the recent update to openssl 1.1, and API_DEPENDS to require it for
>> all ssl-using packages, there is a report of mysql 5.5 failing to build
>> on netbsd-8.  It also fails in mef@'s 9.0_RC1 bulk build:
>> 
>>   http://www.ki.nu/pkgsrc/reports/current/NetBSD-9.0_RC1/20200115.1134/mysql-client-5.5.62nb3/build.log
>> 
>> due to an apparent withdrawal of a DH structure.
>
> mysql-5.5 is long EOL, and should be removed.

Sure - we can propose that once it isn't used by default...

>> Does anyone have a sense of how many failures were caused by openssl,
>> and prognosis for addressing?
>
> None of note for our builds, hence using openssl-1.1.1 in our 2019Q4
> release.

Do you mean, "other than mysql-5.5 which is EOL and for which there is
no reason to use, only a few things broke, and they are all unmaintained
and deletion candidates"?

I am just trying to understand whether other than fixing the default,
there is a tiny maybe-problem or something larger.   It does seem that
mysql-using packages in pkgsrc are now broken, which isn't ok.

>> Sort of separately, mysql seems strange as a project in that we don't
>> have a numbe of the newer versions, and nobody seems to want them,
>> perhaps because of too-radical changes.  We do have a mariadb version,
>> and there are percona versions in the joyent fork of pkgsrc.  Can anyone
>> explain the big picture, and what we should be doing vs where we are?
>
> Mostly lack of those elusive circular tuits on my part, Percona is
> much preferred IMO, I try to keep it reasonably updated in our
> repository and it works quite well as a cluster, e.g.:
>
>   https://github.com/joyent/pkgsrc/wiki/use:percona-cluster
>
> We haven't kept MariaDB as current, but it's popular on other OS and
> it would be good if someone was interested in it on our side.

Thanks for the characterizations.

> The complicated part is bringing over the changes from our pbulkmulti
> branch so that users aren't forced into a single mysqlclient
> implementation.

I don't follow this.  Right now pkgsrc users have to choose between
mysql 5.5 5.6 or 5.7, and pretty clearly ought to choose 5.7 (from among
those).  Extending that set to add mariadb and percona seems like a good
step forward.   I can see the point that we would like bulk builds for
multiple choices, but I don't see why that should block progress.   Is
that right, or am I missing something?