Re: tests fail on windows with default git settings

> The other failures I see are the following, which I'm just starting to dig
> into:
>
>  26/298 postgresql:recovery / recovery/019_replslot_limit
>             ERROR            43.05s   exit status 2
>  44/298 postgresql:recovery / recovery/027_stream_regress
>             ERROR           383.08s   exit status 1
>  50/298 postgresql:recovery / recovery/035_standby_logical_decoding
>             ERROR           138.06s   exit status 25
>  68/298 postgresql:recovery / recovery/040_standby_failover_slots_sync
>              ERROR           132.87s   exit status 25
> 170/298 postgresql:pg_dump / pg_dump/002_pg_dump
>              ERROR            93.45s   exit status 2
> 233/298 postgresql:bloom / bloom/001_wal
>              ERROR            54.47s   exit status 2
> 236/298 postgresql:subscription / subscription/001_rep_changes
>              ERROR            46.46s   exit status 2
> 246/298 postgresql:subscription / subscription/010_truncate
>             ERROR            47.69s   exit status 2
> 253/298 postgresql:subscription / subscription/013_partition
>              ERROR           125.63s   exit status 25
> 255/298 postgresql:subscription / subscription/022_twophase_cascade
>             ERROR            58.13s   exit status 2
> 257/298 postgresql:subscription / subscription/015_stream
>             ERROR           128.32s   exit status 2
> 262/298 postgresql:subscription / subscription/028_row_filter
>             ERROR            43.14s   exit status 2
> 263/298 postgresql:subscription / subscription/027_nosuperuser
>              ERROR           102.02s   exit status 2
> 269/298 postgresql:subscription / subscription/031_column_list
>              ERROR           123.16s   exit status 2
> 271/298 postgresql:subscription / subscription/032_subscribe_use_index
>              ERROR           139.33s   exit status 2

Hm, it'd be good to see some of errors behind that ([1]).

I suspect it might be related to conflicting ports. I had to use
PG_TEST_USE_UNIX_SOCKETS to avoid random tests from failing:

          # use unix socket to prevent port conflicts
          $env:PG_TEST_USE_UNIX_SOCKETS = 1;
          # otherwise pg_regress insists on creating the directory and does it
          # in a non-existing place, this needs to be fixed :(
          mkdir d:/sockets
          $env:PG_REGRESS_SOCK_DIR = "d:/sockets/"

No, it all seems to be fallout from GSSAPI being included in the build. If I build without that, everything passes. Most of the tests are failing with a "too many clients already" error, but a handful do seem to include GSSAPI auth related errors as well. For example, this is from 

... this is from subscription/001_rep_changes: 

[14:46:57.723](2.318s) ok 11 - check rows on subscriber after table drop from publication
connection error: 'psql: error: connection to server at "127.0.0.1", port 58059 failed: could not initiate GSSAPI security context: No credentials were supplied, or the credentials were unavailable or inaccessible: Credential cache is empty
connection to server at "127.0.0.1", port 58059 failed: FATAL:  sorry, too many clients already'
while running 'psql -XAtq -d port=58059 host=127.0.0.1 dbname='postgres' -f - -v ON_ERROR_STOP=1' at C:/Users/dpage/git/postgresql/src/test/perl/PostgreSQL/Test/Cluster.pm line 2129.
# Postmaster PID for node "publisher" is 14488
### Stopping node "publisher" using mode immediate
# Running: pg_ctl -D C:\Users\dpage\git\postgresql\build/testrun/subscription/001_rep_changes\data/t_001_rep_changes_publisher_data/pgdata -m immediate stop
waiting for server to shut down.... done
server stopped
# No postmaster PID for node "publisher"
# Postmaster PID for node "subscriber" is 15012
### Stopping node "subscriber" using mode immediate
# Running: pg_ctl -D C:\Users\dpage\git\postgresql\build/testrun/subscription/001_rep_changes\data/t_001_rep_changes_subscriber_data/pgdata -m immediate stop
waiting for server to shut down.... done
server stopped
# No postmaster PID for node "subscriber"
[14:46:59.068](1.346s) # Tests were run but no plan was declared and done_testing() was not seen.
[14:46:59.069](0.000s) # Looks like your test exited with 2 just after 11.