I see this consistent failure on trunk, I don't recall seeing before.
Tests began at 2026-06-13 13:34:13 IST
exec.test
==== exec-19.1 exec >> uses O_APPEND FAILED
==== Contents of test case:
# Note that we have to allow for the current contents of the temporary
# file, which is why the result is 14 and not 12
exec /bin/sh -c {for a in 1 2 3; do sleep 1; echo $a; done} >>$tmpfile &
exec /bin/sh -c {for a in 4 5 6; do sleep 1; echo $a >&2; done} 2>>$tmpfile &
exec /bin/sh -c {for a in a b c; do sleep 1; echo $a; done} >>$tmpfile &
exec /bin/sh -c {for a in d e f; do sleep 1; echo $a >&2; done} 2>>$tmpfile &
# The above four shell invocations take about 3 seconds to finish, so allow
# 5s (in case the machine is busy)
after 5000
# Check that no bytes have got lost through mixups with overlapping
# appends, which is only guaranteed to work when we set O_APPEND on the
# file descriptor in the [exec >>...]
file size $tmpfile
---- Result was:
22
---- Result should have been (exact matching):
26
==== exec-19.1 FAILED
|