Arkadiy Vertleyb wrote:
Oh, I haven't realised how big a difference in time between tests can be... It makes sense now.
Yes, the situation regarding my test runs is pretty bad currently: I can't leave the testing machine unattended currently: the old PSU broke, I added a new one, there was a expliosion inside the box (a capacitor, likely), visual inspection didn't reveal anything. I currently have to sit right next to the box in order to be able to extinguish any fires that might break out in case there's another explosion. A clean test run takes a day of CPU time. Nearly-clean runs get triggered by small changes to any one of many core libraries (MPL, type_traits, etc). Tests runs need even more attention because some of the tests don't link for intel-9.0. This results in a busy loop and I have to kill the process manually. Once I can trust the test machine again and the linking problem got fixed, results should be available more frequently again. Regards, m Send instant messages to your online friends http://au.messenger.yahoo.com