bug in the testsuite: doesn't accept freshly started tcp redis server as ready #3
Description
in release 0.48 (which isn't present as tag or release on github, btw), and with a redis server version 6.0.16 the testsuite doesn't run t/09-conn-errors.t properly: after test 14 there's a very long delay (around a minute), and then the remaining tests are all skipped: "redis-server is required for this test".
the root cause is that t/tlib/Test/RedisRunner.pm misidentifies its newly started redis-server (with port) as not ready, waits a while and then kills it (and the rest of the testsuite unfortunately swallows the diagnostic output from redisrunner, so it's a bit hard to find out what happened).
redisrunner looks for this:
if ( $log =~ /The server is now ready to accept connections/
whereas redis 6 logs this if (and only if) started with a tcp port:
26223:C 27 Apr 2023 10:43:46.935 # Redis version=6.0.16, bits=64, commit=00000000, modified=0, pid=26223, just started
....
26223:M 27 Apr 2023 10:43:46.936 * Running mode=standalone, port=38217.
26223:M 27 Apr 2023 10:43:46.936 # Server initialized
...
26223:M 27 Apr 2023 10:43:46.936 * Ready to accept connections
the interesting (well, ugly) aspect is that redis logs a different message if you start it with only a socket:
17369:M 26 Apr 2023 10:35:58.265 * The server is now ready to accept connections at /var/run/redis/redis-server.sock
solution: i'd suggest changing redisrunner to look for /[rR]eady to accept connections/ which covers both.