[Python-checkins] r54805 - python/branches/release25-maint/Lib/test/test_pty.py
barry.warsaw
python-checkins at python.org
Fri Apr 13 18:12:05 CEST 2007
Author: barry.warsaw
Date: Fri Apr 13 18:12:02 2007
New Revision: 54805
Modified:
python/branches/release25-maint/Lib/test/test_pty.py
Log:
Add code to read from master_fd in the parent, breaking when we get an OSError
(EIO can occur on Linux) or there's no more data to read. Without this,
test_pty.py can hang on the waitpid() because the child is blocking on the
stdout write. This will definitely happen on Mac OS X and could potentially
happen on other platforms. See the comment for details.
Modified: python/branches/release25-maint/Lib/test/test_pty.py
==============================================================================
--- python/branches/release25-maint/Lib/test/test_pty.py (original)
+++ python/branches/release25-maint/Lib/test/test_pty.py Fri Apr 13 18:12:02 2007
@@ -115,6 +115,24 @@
os._exit(4)
else:
debug("Waiting for child (%d) to finish."%pid)
+ # In verbose mode, we have to consume the debug output from the child or
+ # the child will block, causing this test to hang in the parent's
+ # waitpid() call. The child blocks after a platform-dependent amount of
+ # data is written to its fd. On Linux 2.6, it's 4000 bytes and the child
+ # won't block, but on OS X even the small writes in the child above will
+ # block it. Also on Linux, the read() will throw an OSError (input/output
+ # error) when it tries to read past the end of the buffer but the child's
+ # already exited, so catch and discard those exceptions. It's not worth
+ # checking for EIO.
+ while True:
+ try:
+ data = os.read(master_fd, 80)
+ except OSError:
+ break
+ if not data:
+ break
+ sys.stdout.write(data.replace('\r\n', '\n'))
+
##line = os.read(master_fd, 80)
##lines = line.replace('\r\n', '\n').split('\n')
##if False and lines != ['In child, calling os.setsid()',
More information about the Python-checkins
mailing list