On Tue, 17 Feb 2026 at 18:57, Ilya Maximets <[email protected]> wrote:
> On 2/11/26 4:09 PM, Matteo Perin via dev wrote: > > Tests > > OVS datapath key parsing and formatting - valid forms > > OVS datapath actions parsing and formatting - userdata overflow > > sometimes failed with stderr output: "sed: couldn't write X > > items to stdout: Broken pipe" and "cat: actions.txt: EPIPE: > > Broken pipe". > > > > This occurred due to a race condition where the python test script > > (test-dpparse.py) would finish processing and close its stdin before > > the pipeline (cat | sed | sed) completed writing all its output. When > > this happened, both cat and sed received SIGPIPE and wrote error > > messages to stderr, causing the AT_CHECK macro to fail the test. > > Hi, Matteo. It should not be possible for the loop in the script to finish > before the stdin is closed, which will happen only after sed is done with > the > output. The only case where this would happen is if there was an exception > in the code, but in this case the test would fail anyway, as no output is > expected. Or am I missing something? > > Best regards, Ilya Maximets. Hi Ilya, No, you are totally right. I was so misled by the multiple piping, the python script execution flow and the resulting log output that I convinced myself the race condition was the correct explanation. After your comment I looked into it more accurately and, as you correctly suspected, the failure was indeed simply caused by a (silent) exception in the script, in this case a missing dependency in the build env I was using. Sorry for the confusion, please disregard this patch as it does hide a possible underlying problem with the script, after all. Best Regards, Matteo _______________________________________________ dev mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-dev
