You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The signal handler should collect the signals and waitpid should deliver them. The signal may occur before subprocess.Popen returns, that is, the child (from pksubprocess_test) executes before the parent process has gotten the pid from subprocess. This occurs in parallel test runs. Simplifying the signal handler is better anyway. This will require a change to the way the non-recursive waitpid works. It's ok to poll. The key is making it fast enough that higher level processes (e.g. sirepo job supervisor) do not send their secondary signal before the signal gets to be cascaded. This should be fine. The poll loop can actually be structured so that the check for signals happens every .1 and the ps calls happen every 1 second.
The text was updated successfully, but these errors were encountered:
…tion,
which only shows up when running parallel tests. #554 is not fixed by this
commit.
test.sh also now prints the pid/time, which is helpful for debugging.
- Added quest.Attr.IS_SINGLETON to enable http.Session
- documented quest.Attr.
- Fixed pksubprocess_test (hopefully) after discovering #544 race condition,
which only shows up when running parallel tests. #554 is not fixed by this
commit.
- test.sh also now prints the pid/time, which is helpful for debugging.
The signal handler should collect the signals and waitpid should deliver them. The signal may occur before subprocess.Popen returns, that is, the child (from pksubprocess_test) executes before the parent process has gotten the pid from subprocess. This occurs in parallel test runs. Simplifying the signal handler is better anyway. This will require a change to the way the non-recursive waitpid works. It's ok to poll. The key is making it fast enough that higher level processes (e.g. sirepo job supervisor) do not send their secondary signal before the signal gets to be cascaded. This should be fine. The poll loop can actually be structured so that the check for signals happens every .1 and the ps calls happen every 1 second.
The text was updated successfully, but these errors were encountered: