[PATCH] WIP: new insert tests for failing indexing.

[David Bremner]

The second one fails because the initial write lock fails.
I guess this is maybe a doc bug.
---
 test/T070-insert.sh | 21 +++++++++++++++++++++
 1 file changed, 21 insertions(+)

This is what I came up with when trying to induce indexing
failure. The best I could do was make the initial notmuch open fail.
This means my second test is broken.  When I was writing the commit
message I was thinking the man page was therefore wrong, but actually
it's ok.

I did wonder if it might be a good idea to move the database open
after the message is written to disk, is we want to maximize the
chance of insert --keep "succeeding".

diff --git a/test/T070-insert.sh b/test/T070-insert.sh
index aacc643..e7d0ae1 100755
--- a/test/T070-insert.sh
+++ b/test/T070-insert.sh
@@ -183,4 +183,25 @@ test_expect_code 1 "Invalid tags set exit code" \
 
 notmuch config set new.tags $OLDCONFIG
 
+
+gen_insert_msg
+test_python <<EOF &
+import notmuch, time, os
+with open("write-locker.pid","w") as pidfile:
+    pidfile.write(str(os.getpid()))
+db = notmuch.Database(mode=notmuch.Database.MODE.READ_WRITE)
+time.sleep(120)
+EOF
+
+# give time for the background process to get started
+sleep 1
+
+test_expect_code 1 "Non-zero exit code due to write lock" \
+    'notmuch insert < "$gen_msg_filename" 2>&1'
+
+test_expect_code 0 "Zero exit code due to --keep flag" \
+    'notmuch insert --keep < "$gen_msg_filename" 2>&1'
+
+kill $(cat ${TMP_DIRECTORY}/write-locker.pid) >/dev/null 2>&1
+
 test_done
-- 
2.1.0