[MacPorts] #52144: mariadb: support for mpkg / mdmg

Craig Treleaven ctreleaven at macports.org
Mon Sep 5 13:34:42 PDT 2016


> On Sep 5, 2016, at 4:09 PM, Bradley Giesbrecht <pixilla at macports.org> wrote:
> 
>> On Sep 5, 2016, at 8:03 AM, Lawrence Velázquez <larryv at macports.org> wrote:
>> 
>>> On Sep 5, 2016, at 8:42 AM, Craig Treleaven <ctreleaven at macports.org> wrote:
>>> 
>>> Not to belabour the issue, but should it not be the impact to port users that determines whether a change is “minor” or not?
>> 
>> I believe the "minorness" of the change is wholly up to the maintainer.
>> 
>>> The number of lines, by itself, doesn’t necessarily determine that impact. For example, a 1 or 2 line change in one of the database ports might make a new database engine the default.
>> 
>> It is certainly true that a small change with great impact is not minor, but a large change with little impact is also not minor. As an extreme example, I would not appreciate a commit to one of my ports that had no impact on the installation yet completely rearranged the portfile. I'd have to waste time reading and understanding the committer's code, looking for edge cases and failure modes, reworking local commits that no longer apply, etc.
>> 
>> (This situation can already happen via timeout, but in that case there is a clear, objective policy that maintainers implicitly agree to when they take up maintainership.)
>> 
>> My rule of thumb is that fixing typos and broken builds is almost always okay under openmaintainer. Many maintainers also permit minor version bumps and bug fixes, but some don't. In all cases, it's safest to wait out the 72 hours.
> 
> I was going to add to the ticket but maybe this is a better place to discuss for now.
> 
> I have attempted to keep the mysql ports similar to make maintaining them easier.
> 
> I have a few questions regarding these changes.
> 
> 1. Will "port load mariadb-server" work?

Good point, I don’t know.  I’ll try it.

BTW, this afternoon I noticed that I’m not respecting the “startupitem.install” setting.  I’ll change the patch so if the user has modified macports.conf with ‘startupitem.install no’, we don’t put anything in /Library/LaunchDaemons.

> 2. Has this been tested on older systems?

Not yet.  It has been a lot of time and work to get this far.  What is the earliest OS that you test with the server ports?  The new version of my MythTV ports are not going to support anything before 10.9!  The best I can do is 10.6 on X86_64.

> 3. Can the pkg run concurrently with the port? I assume pkg installs into $prefix, am I right?

Correct, the pkg installs in the same prefix it was built in.  For development, I build mpkgs on my main system (10.10) and test installation on a VM (10.11).  (Later, I’ll build the mpkg for distribution on a VM with a non-default prefix.)  

> 4. Is there any reason the same changes would not work for the mysql*-server, mariadb*-server and persona-server ports?

I know almost nothing about percona but I think it ought to work with the different versions of mysql and mariadb.  The launchd plist was introduced in MySQL 5.7.8.    

Craig


More information about the macports-dev mailing list