Monday, February 28, 2011

OSX 10.6.5 update breaks apachectl script

If you're wanting to manipulate Apache via the apachectl script, OSX 10.6.5 apparently changed a line in the script in such a fashion that it no longer functions.

In the interest of brevity I simply recommend what others have suggested, change the line:

ULIMIT_MAX_FILES="ulimit -S -n 'ulimit -H -n'"

To simply be:

ULIMIT_MAX_FILES=""

If you need greater control of the directive, knock yourself out :)

Friday, February 25, 2011

Parallels performance tests

I spent part of my day evaluating performance for Windows 7 Ultimate 64-bit under Parallels on Mac OSX Snow Leopard.

Disheartening is the way I would put it. Thinks work well for the most part, but in other ways it falls short.

I really was expecting better results than a 50% performance hit compiling with VS 2010 Professional. Clean, Rebuild of 27 projects as Win32 Debug in the solution, roughly 400,000 lines of C++ code = 4 minutes 18 seconds on average. Under Windows 7 native that time was cut to an average of 1 minute 50 seconds. No concurrent builds allowed.

The virtual machine running off the bootcamp partition is allocated 6 processors and 8GB of memory. The native machine has 12 cores with 24 hardware threads (Parallels doesn't seem to provide hardware threads to VM as I allocate 6 cores and 6 procs show up in the Task Manager [you would think 12 would.]) This is why I expected a performance drop, but not a more than 100% improvement under the native OS.

Anyhow, still very impressive software, and it is nice to have OSX and Win 7 running side by side. I'm probably doing something wrong anyhow (historical precedent would suggest...)

Thursday, February 24, 2011

Parallels and PRL_ERR_DISK_GPT_MBR_NOT_EQUAL

This should help out some poor b**tard like me... I installed Parallels today to give it a spin because I heard that you can now run a virtual machine directly off your Boot Camp partition (ergo I can run pure Windows like I need to in order to properly test software, and I can run the very same partition as a virtual machine in OSX.)

Well, as usual, I ran into some hiccups; however, this time the hiccups were from something I'd done (but still shouldn't have been hiccups really.

For some reason, when I tried to get Parallels to use my Boot Camp partition, I was getting a 'cannot configure the boot camp partition' error and when I looked at the details I received a very mysterious PRL_ERR_DISK_GPT_MBR_NOT_EQUAL error.

Looking at this error suggests there some sort of problem with the GPT and the MBR (if you don't know what these are you aren't ready for this post, wikipedia is your friend) where something was not what someone expected.

Well, I know that after creating my Boot Camp partition in OSX that I'd messed around with that partition space in Windows 7 to create and extended partition by shrinking the Boot Camp partition. OSX didn't seem to care about this and was quite happy to mount the partitions for read only access. Well, it turns out that Parallels cares about this a lot and when it looks at the GPT to see what the MBR should be it throws up because the GPT states that the end of the Boot Camp partition is something it is not.

Sooooo... Long story short, it turns out I'm lucky that I'm using reFIT to handle my multi-boot system because when reFIT starts up you can run a little utility called gptsync (choose 'run partition tool' from the reFIT boot menu) which apparently will square up what is stated in the GPT with what is stated in the MBR (technically it modifies the MBR so be careful.)

BTW, gptsync doesn't handle extended partitions so I had to move all of that data onto the main Boot Camp partition and delete the partition then re-extend the Boot Camp partition over that space so it resembled what OSX remembered from earlier - then I could run gptsync.

Booting OSX and running Parallels at this point worked fine. As an aside, I was wondering if this was a Parallels limitation but it turned out to be a problem (again with virtually no intelligent description) for VMWare's Fusion 3.1 as well.

Hope this helps somebody! :)

UPDATE: Don't forget, both in VMWare Fusion and Parallels, to delete any previous virtual machine you had tried to make using that Boot Camp partition because it has erroneous GPT data - start with a fresh virtual machine.

Saturday, February 19, 2011

Home and End keys in most editors for OSX

http://www.starryhope.com/tech/apple/2006/keyfixer/

God bless that guy, I really didn't want to create a custom keybinding dictionary. Works for my Ergonomic 7000. 'Nuff said.

Wednesday, February 16, 2011

If your non-Apple keyboard won't eject using F12...

...make sure that if your keyboard preferences (either in Apple's preferences or a preference list created by keyboard install software on your machine) offer a keyboard type of 'Generic', try setting it to this value.

My new Microsoft Ergonomic 7000 Wireless was set to ANSI for some reason after installation of the software on OSX and F12 would bring up dashboard (like it should) but would not trigger the open/close tray command when depressed longer (as it was supposed to.)

Switching the setting from 'Ansi' to 'Generic' solved this for me (for this keyboard the value was under 'Options' in the system preferences->Microsoft Keyboard.)

Hope this helps someone someday :).

Monday, February 7, 2011

Bought a new crApple Mac Pro for work...

...the other day.

Un-freaking-beleivable how ridiculously stupidly overpriced the thing is. Oh, don't get me wrong, it's sexy, sleek, and oh so very shiny with 12 cores, and multiple terabytes of drives (et cetera); but, how on earth are there people so stupid that they actually buy memory from crApple? 24GB of memory (you start with 6GB) from Apple is more than $2100. You can buy 24GB from Crucial for about $420 and then you get to keep 2 of the 1GB sticks that crApple gives you by default for a total of 26GB, plus you get save $1700 (think of all the iPads, iPhones, iPods, iTouches, iWhatevers you could buy with that.)

I love the Woz, hate the Jobs...