After r50784 many storage tests are failing on the SnowLeopard leaks bot. See <http://build.webkit.org/results/SnowLeopard%20Intel%20Leaks/r50784%20%282092%29/results.html> for the first set of failing results. The storage tests have been failing consistently since then.
Created attachment 43082 [details] Patch
Comment on attachment 43082 [details] Patch After talking to bdash, we decided this wasn't particularly useful...and that we just need better error handling in general.
It turns out that the cause of the failure here was the hard disk filling up. But it sure was hard to tell based on what these tests outputted and even the error messages in stderr. Lets try to make the error handling more robust. Dumi, can you take this on?