One of my two computers just crashed, and one of your Human Proteome Folding jobs appears to be responsible.

Its BOINC completed some job or another, started a Rosetta task, and the Rosetta promptly crashed with some sort of "C++ runtime error". That machine's BOINC moved on to a second Rosetta job, and within a few minutes it crashed in the same manner. Then the following all crashed within a few minutes of starting: an FAAH job, another FAAH job, and another HPF job. Then OTHER applications started crashing with the same error message. Just about everything crashed, including Explorer, and Explorer restarted. The computer had become unusable however -- running anything would cause it to immediately crash with the same error message. Not a normal "This program has encountered a problem..." but some sort of C++ runtime error message, with only slight variations.

I had a task manager running, and noticed that everything that crashed first leaked -- no, hemmorhaged memory, bloating to over 1GB process size(!) before crashing.

The computer was working fine right up until that first Rosetta job crash. It was working fine yesterday at this time, and the day before. I hadn't changed anything. As far as I can tell, the first Rosetta crash started some kind of chain reaction that corrupted the system in some way.

In other words, Rosetta has a severe bug, and work units that have some particular X-factor in them can trigger this bug, which will cause Rosetta to crash and corrupt the system.

The truly amazing thing is that this isn't some poky Windows ME box. It's running XP Pro, a fully 32 bit protected-mode operating system that is supposed to prevent application faults from corrupting the entire system. I've never seen one crash before other than due to kernel-mode failures, usually in a video driver, when the crash takes the form of a blue screen. This wasn't a BSOD, yet the system did become corrupted, and it started with a Rosetta binary running and then crashing.

Please look into this. Anything that can take down a protected-mode operating system without having to run in kernel mode is a serious problem indeed, and might even be abused. There's clearly also a bug in WinXP for this to even be possible. With any luck, MS will fix it soon, since it has security implications (obviously it enables an unprivileged user to launch a denial of service attack on an XP box, and it might enable one to gain privilege depending on the exact nature of the bug). But until then it behooves WCG to try not to crash members' computers; widespread occurrences of what just happened to me will discourage participation, which I doubt you would enjoy. Even if MS fixes *their* bug, Rosetta crashing sometimes will rob users of points, WCG of usable HPF results, and the whole community of CPU cycles that go down the drain.

Oh, and the "runtime error" thing steals focus, which is annoying.