Re: [gentoo-alt] Prefixed MySQL on MacOSX-intel

[Fabian Groffen]

Hi Ramon,

On 08-09-2008 16:21:56 +0200, Ramon van Alteren wrote:
> Because I need / want to develop some stuff against MySQL-5.1 on my
> Macbook I started porting the mysql-ebuild to my prefix environment
> using ecopy and friends.
> In addition I need dev-python/mysql-python and sqlalchemy with the mysql
> useflag turned on.
>
> I've ecopy'd:
>
> dev-db/mysql-community-5.1.21_beta
> dev-python/mysql-python-1.2.2-r1
> dev-db/mysql-init-scripts-1.2
> virtual/mysql-5.1
>
> Helpful bugs are:
>
> Porting in general:
> https://bugs.gentoo.org/attachment.cgi?id=164488
> (Sorry forgot to bookmark the bug and am too lazy to search my IRC logs
> for it )

http://www.gentoo.org/proj/en/gentoo-alt/prefix/ecopy.xml (off the top
of my head)

> mysql-python:
> (also contains description on how to get mysql running/ecopy'd)
> http://bugs.gentoo.org/show_bug.cgi?id=220753
>
> init-scripting:
> http://bugs.gentoo.org/show_bug.cgi?id=196294

Not a necessity I think, but would be really nice to have.

> The result is that I have a working ebuild for
> mysql-community-5.1.21_beta which I emerged with use-flags: ssl
>
> As soon as I add innodb as use-flag innodb cmake gets pulled in as a
> dependency which fails to build on macosx-x86, possibly due to
> https://bugs.gentoo.org/show_bug.cgi?id=236740
>
> I have not yet had the time to investigate any further into this.

Ahhhh, painful.  Does a lower version work?

> emerge --config fails due to an unknown option used when starting the
> server (--skip-ndbcluster) which is added in the mysql.eclass for all
> server above version 4.1.3, if I remove that there's an unknown option
> for --skip-innodb, seems like the eclass might need a small revamp to
> deal with the new plugin setup in 5.1

Hmmm...

> Caveats:
>
> install-db scripting needs to be run by hand.
> no innodb support
> no working init-script
> ownership of the logfiles need to be reset by hand
>
> I have the following questions:
>
>  * What testing if any is normal for this kind of stuff ? I'm looking
> into getting the test-suite to run on the installed bins, is that
> mandatory for prefix ?

Running the testsuite is of course very helpful.  However, the best
tests come from advanced users that know how the product works, and
hence can quickly do a sanity test to see if the product behaves on a
few core tasks.

>  * What is the current idea of init-scripts for these type of services,
> should the init-scripts be patched to work on alt-platformX ?

They require a lot of work.  I'd love to have them, but for now we have
to ignore them, until we have a working infrastructure for them (and
good plans on how to "port" them).

>  * How to get these ebuilds into prefix, open a bug ?

Bugs work usually best, so anyone on the team can pick them up.  What we
like to know is:
- which changes did you make *after* ecopy, preferably using a diff -u
- which patches you added, preferably attached uncompressed
- if it runs, and if so, as expected

>  * What about the eclass updates that are necessary (this most likely
> affects current "normal" gentoo tree as well ?)

I have to look into the details for this one, but I think this is indeed
a gentoo-x86 problem, judging from your description.  That would
preferably solved first there, such that it automagically gets into
Prefix.

>  * Did I miss anything that needs to be done ?

I can't instantly name one, so I guess not!

Thanks for your detailed explanation.  I hope we can incorporate your
efforts in the tree soon.


-- 
Fabian Groffen
Gentoo on a different level