In August Productions

The latest version of Sun Studio, which is a terrible name for something which is a compiler and not an IDE, which has now been re-titled Solaris Studio in an attempt to underscore even more its general uselessness, has added a new antifeature to its installer.

It requires X.

You know one of the things that makes MySQL better than Oracle? You can install it in 5 minutes and you don't have to launch a Java GUI.

You know when it makes sense to launch a Java GUI to install server software?

NEVER

EPIC FAIL 

When you rely on something like Jenkins to help manage your project, if you don't have a feature regularly tested in it, that feature often bitrots pretty quickly. This was the case for our Windows compile of libdrizzle. It worked months ago, but then some folks tried it recently and, well - not so much with the working. That meant I had to do three things this weekend:

1. Get the Windows machine that was sitting at my old office powered off and actually plug it in in my new office
2. Fix the build
3. Connect that machine to Jenkins

I'm happy to report all three are done, and we now have a Jenkins slave building libdrizzle on every push. I believe I should also be able to build the client programs, so I'll add them soon - and then it's only a few steps away from having Jenkins produce an msi or something that people can download.

In case you're wondering, we are currently aiming for Windows7/Windows 2008 Server. 

Also, now that we have the Jenkins slave, I can get the same thing done for libmemcached.

I've been remiss in blogging about this - been on my todo list for a while, but you know, coding is less exciting than blogging.

But then this morning I was reading Shane's post about Quickly templates for Vala and one he hacked up for C, which has spurred me on to actually get off my butt and write a blog post - since I've had a quickly template which supports C and C++ for months now. Sigh. Sorry.

Anyhow - the template a part of the pandora-build project, and can be installed with:

 add-apt-repository ppa:drizzle-developers/ppa

 apt-get install pandora-build

At that point:

 quickly create pandora-build foo application

or

 quickly create pandora-build foo library

Should work. There are two different sub-commands for making a lib or an app, since those are a slightly different layout. All of the projects created are fully pandora-build enabled. There is a quickly upgrade command for upgrading to the latest installed pandora-build for the project, and a quickly add class command which will create new stub class files for you and add them to the build system. The stub project should work immediately with autoreconf -i ; ./configure ; make distcheck

There are also pandora-build-c templates if you want C rather than C++, but I just gave them a test and there is a bug that I  fixed in the c++ versions but didn't carry over to the C templates. I'll try to get that sorted real soon now.

Since this is all driven by my managing of the build systems withing Drizzle, Gearman and libmemcached, I keep adding bits in from our standards there - so I've got adding in support for gtest test suites and Sphinx documentation on my todo list. I also need to add support for spinning up debs, but I'm a little at odds with how that should work as I'm not so much a believer in debian packaging in upstream sources... so I'll have to battle with that later.

Update: Turns out I suck - I didn't upload to Maverick - fixing. 

Update: Ok. C-template problem fixed. v0.156. New packages uploaded to the drizzle-developers PPA. 

I just made a patch to the OpenStack tarmac instance which was merged back into upstream trunk. There is now an option (which we are using) which removes the requirement for a separate commit message to be present on the merge proposal before an approved proposal can be merged. It will still use an explicit commit message if it's there, but if it's not, it will happily use your merge proposal description as a commit message.

I do implore you though - if you're the type of person who likes to write 10 page merge proposal descriptions, please also add a short commit message.

Thanks to Jobin's work with mingw and getting libdrizzle to compile on Windows at all, I have been able to get it working in Visual Studio natively. The code is in trunk now.

The approach I took, which is how I'm going to approach Windows and Visual Studio for all of our stuff, is to not worry with analogues to things like configure on Windows. Windows is a very different platform from Linux, and there is no need to attempt to duplicate Linux process there. To that end, the goal at least for now will be static VS Solution files and a set of instructions of how to get depends installed so that the Solution can find them. 

I'm excited to start poking at Garrett Serack's CoApp Project, which has some tools do do tracing of things like make to help with the initial project creation... and I fully expect to take advantage of his work in terms of creating installable packages.

Jobin has started working on Drizzle itself and has the drizzle client working so far.