diff options
author | Tom Lane <tgl@sss.pgh.pa.us> | 2019-08-04 13:07:12 -0400 |
---|---|---|
committer | Tom Lane <tgl@sss.pgh.pa.us> | 2019-08-04 13:07:12 -0400 |
commit | 45d6789e78dc2997292a528e2538fa496485b85f (patch) | |
tree | 9061f7a9b3862426650238943a9691be8a83c359 /src/backend/utils/adt/jsonb_util.c | |
parent | e3967a16d3a0b9c1c6ca20edccc2696e313127e9 (diff) | |
download | postgresql-45d6789e78dc2997292a528e2538fa496485b85f.tar.gz postgresql-45d6789e78dc2997292a528e2538fa496485b85f.zip |
Avoid picking already-bound TCP ports in kerberos and ldap test suites.
src/test/kerberos and src/test/ldap need to run a private authentication
server of the relevant type, for which they need a free TCP port.
They were just picking a random port number in 48K-64K, which works
except when something's already using the particular port. Notably,
the probability of failure rises dramatically if one simply runs those
tests in a tight loop, because each test cycle leaves behind a bunch of
high ports that are transiently in TIME_WAIT state.
To fix, split out the code that PostgresNode.pm already had for
identifying a free TCP port number, so that it can be invoked to choose
a port for the KDC or LDAP server. This isn't 100% bulletproof, since
conceivably something else on the machine could grab the port between
the time we check and the time we actually start the server. But that's
a pretty short window, so in practice this should be good enough.
Back-patch to v11 where these test suites were added.
Patch by me, reviewed by Andrew Dunstan.
Discussion: https://postgr.es/m/3397.1564872168@sss.pgh.pa.us
Diffstat (limited to 'src/backend/utils/adt/jsonb_util.c')
0 files changed, 0 insertions, 0 deletions