There's only one regression test failure left on NetBSD (all versions and architectures), and it's one that can actually affect real-world scripts. It would be really nice if we could fix that before the next release.
test io begins at 2026-04-02+00:08:10
io.sh[653]: FAIL: 'read' hangs on EOF without final linefeed when reading from FIFO
test io failed at 2026-04-02+00:08:16 with exit code 1 [ 175 tests 1 error ]
It's this test from 9ba2c2e:
|
# On unpatched ksh on macOS, 'read' used to block when reading from a FIFO and there was no final newline. |
|
if mkfifo "$tmp/fifo_no_lf" |
|
then trap 'sleep_pid=0; kill "$ksh_pid"; err_exit "'\''read'\'' hangs on EOF without final linefeed when reading from FIFO"' TERM |
|
(sleep 1; kill "$$") & |
|
sleep_pid=$! |
|
"$SHELL" -c 'print -n foo >$0 & while read f; do :; done <$0' "$tmp/fifo_no_lf" & |
|
ksh_pid=$! |
|
wait "$ksh_pid" |
|
trap - TERM |
|
((sleep_pid)) && kill "$sleep_pid" |
|
else err_exit "mkfifo failed; cannot test reading from FIFO" |
|
fi |
There's only one regression test failure left on NetBSD (all versions and architectures), and it's one that can actually affect real-world scripts. It would be really nice if we could fix that before the next release.
It's this test from 9ba2c2e:
ksh/src/cmd/ksh93/tests/io.sh
Lines 646 to 657 in d745cca