I noticed our cruisecontrolled build of the cpp trunk was blocked this
morning. It turned out to be a test that failed to complete (perftest).
I couldn't reproduce it when running make check manually on the same
box, but I saw the same thing again on the first attempt to restart
cruisecontrol (a second attempt passed).
Its fairly hard to reproduce reliably, at least for me. Am investigating
further, so far without a great deal of success (unfortunately any
logging or debug seems to mask the problem), but thought it was worth
mentioning.
- [c++] warning: race causing test to hang occasionally Gordon Sim
-