https://bz.apache.org/SpamAssassin/show_bug.cgi?id=8040
--- Comment #5 from Sidney Markowitz <sid...@sidney.com> --- I have a possible race condition as a cause, after poring over the code and logs of passed tests. The test starts spamd with -D option, with a test plugin defined. Stderr is sent to a log file. The plugin has a dbg line that prints each time it is fired, with a counter in the output line. The test repeatedly calls spamc with mail that should trigger the plugin, than scans the spamd stderr log file, looking for the dbg line with the correct counter value. Since spamd is running as a separate process with the stderr log file always open and being written to, there is nothing to ensure that it has flushed the output buffer after writing that dbg line before the test process reads the stderr log file. I don't now how we would ensure that without adding too much complication to what is supposed to be just a test of the plugin framework. Since I have only seen an error once out of thousands of runs of this test on various test machines, I'm inclined to "fix" this by adding a diag note to the test saying that a race condition can cause a rare test failure that can be ignored. -- You are receiving this mail because: You are the assignee for the bug.
