On Sun, May 18, 2014 at 12:47 PM, Patrice Dumas <pertu...@free.fr> wrote: > Hello, > > With info tests, when doing parallel tests, with > make -j8 check > for example, there may be some failures that do not happen for > sequantial tests. It is not that much reproducible. There is something > in the log that could point to the issue though: > > FAIL: t/body-start.sh > ===================== > > t/Init-intera.inc: line 21: /dev/pts/36: No such file or directory > > The test that fail is not always the same, bug the error message is the > same.
I can't reproduce this. I've added debugging print statements to pseudotty.c, which will be recorded in the *.log files. If you could try it again, you could have another look to see if this gives us any more idea what is causing this. Different test runs could be killing each others pseudotty instances, but I can't see how this could happen. Or different tests could be ending up with the same pts, but all tests I did make this unlikely (every call to getpt() creates a new pts file in /dev/pts/), and even if this did happen, it shouldn't matter. Or pseudotty could be failing and not opening (or closing) the slave file - but this would have to be after it had printed the name of the file to stdout, considering your error message above.