http://www.slony.info/bugzilla/show_bug.cgi?id=156

--- Comment #6 from Christopher Browne <cbbrowne at ca.afilias.info> 2010-11-04 09:13:34 PDT ---
(In reply to comment #5)
> The latest version of the patch seems to introduce the following compiler
> warning
> 
> slon.c: In function ‘SlonMain’:
> slon.c:460: warning: comparison between pointer and integer
> 
> from the line 
> if (PQgetvalue(res, 0, 0) == 'f') {
>                 slon_log(SLON_FATAL, 
>                  "slon_node_health_check
> 
> I am surprised that your testing was able to get this warning to show up.
> 
> PQgetvalue(..) will return  a char * to a null terminated string not a char.
> strncmp is probably what we want to be using.
> 
> 2.  If I modify the code to work around that issue this is all I see when I see
> in the slon output.
> 
> 2010-11-03 10:19:30 EDTCONFIG main: local node id = 1
> 2010-11-03 10:19:30 EDTFATAL  could not call slon_node_health_check() -
> 2010-11-03 10:19:30 EDTCONFIG slon: child terminated status: 9; pid: 31474,
> current worker pid: 31474
> 2010-11-03 10:19:30 EDTINFO   slon: done
> 2010-11-03 10:19:30 EDTINFO   slon: exit(0)
> 
> 
> I'm not seeing the details. (ie the specific table that is the issue)

Right you are...

I have changed it to 
    if (*(PQgetvalue(res, 0, 0)) == 'f') {

which is consistent with what is done elsewhere:

postgres at cbbrowne [12:08:05] [~/slony1-engine.github/src/slon] [bug156 *]
-> % grep -n PQgetvalue *.c | grep "*" | grep "'"
remote_worker.c:2661:           if (*(PQgetvalue(res1, 0, 0)) == 't')
slon.c:460:                             if (*(PQgetvalue(res, 0, 0)) == 'f') {
slon.c:550:             int                     no_active = (*PQgetvalue(res,
i, 1) == 't') ? 1 : 0;


(Of course, the second instance is for this patch! :-))

Here is a self-contained test of this, run against a node that was initialized
and had tables added in.  I mess with it by breaking sl_table.tab_reloid
values.

postgres at cbbrowne [12:10:00] [~/slony1-engine.github/src/slon] [bug156 *]
-> % psql -d slonyregress1 -c "update _slony_regress1.sl_table set
tab_reloid=tab_id;"
UPDATE 4
postgres at cbbrowne [12:10:47] [~/slony1-engine.github/src/slon] [bug156 *]
-> % slon slony_regress1 'dbname=slonyregress1' > /tmp/stdout.txt 2>
/tmp/stderr.txt
postgres at cbbrowne [12:10:55] [~/slony1-engine.github/src/slon] [bug156 *]
-> % cat /tmp/stdout.txt
2010-11-04 12:10:55 EDTCONFIG main: slon version 2.0.5 starting up
2010-11-04 12:10:55 EDTINFO   slon: watchdog process started
2010-11-04 12:10:55 EDTCONFIG slon: watchdog ready - pid = 29512
2010-11-04 12:10:55 EDTCONFIG slon: worker process created - pid = 29513
2010-11-04 12:10:55 EDTCONFIG main: Integer option vac_frequency = 3
2010-11-04 12:10:55 EDTCONFIG main: Integer option log_level = 0
2010-11-04 12:10:55 EDTCONFIG main: Integer option sync_interval = 2000
2010-11-04 12:10:55 EDTCONFIG main: Integer option sync_interval_timeout =
10000
2010-11-04 12:10:55 EDTCONFIG main: Integer option sync_group_maxsize = 20
2010-11-04 12:10:55 EDTCONFIG main: Integer option desired_sync_time = 60000
2010-11-04 12:10:55 EDTCONFIG main: Integer option syslog = 0
2010-11-04 12:10:55 EDTCONFIG main: Integer option quit_sync_provider = 0
2010-11-04 12:10:55 EDTCONFIG main: Integer option quit_sync_finalsync = 0
2010-11-04 12:10:55 EDTCONFIG main: Integer option sync_max_rowsize = 8192
2010-11-04 12:10:55 EDTCONFIG main: Integer option sync_max_largemem = 5242880
2010-11-04 12:10:55 EDTCONFIG main: Integer option remote_listen_timeout = 300
2010-11-04 12:10:55 EDTCONFIG main: Boolean option log_pid = 0
2010-11-04 12:10:55 EDTCONFIG main: Boolean option log_timestamp = 1
2010-11-04 12:10:55 EDTCONFIG main: Boolean option cleanup_deletelogs = 0
2010-11-04 12:10:55 EDTCONFIG main: Real option real_placeholder = 0.000000
2010-11-04 12:10:55 EDTCONFIG main: String option cluster_name = slony_regress1
2010-11-04 12:10:55 EDTCONFIG main: String option conn_info =
dbname=slonyregress1
2010-11-04 12:10:55 EDTCONFIG main: String option pid_file = [NULL]
2010-11-04 12:10:55 EDTCONFIG main: String option log_timestamp_format =
%Y-%m-%d %H:%M:%S %Z
2010-11-04 12:10:55 EDTCONFIG main: String option archive_dir = [NULL]
2010-11-04 12:10:55 EDTCONFIG main: String option sql_on_connection = [NULL]
2010-11-04 12:10:55 EDTCONFIG main: String option lag_interval = [NULL]
2010-11-04 12:10:55 EDTCONFIG main: String option command_on_logarchive =
[NULL]
2010-11-04 12:10:55 EDTCONFIG main: String option syslog_facility = LOCAL0
2010-11-04 12:10:55 EDTCONFIG main: String option syslog_ident = slon
2010-11-04 12:10:55 EDTCONFIG main: String option cleanup_interval = 10 minutes
2010-11-04 12:10:55 EDTCONFIG main: local node id = 1
2010-11-04 12:10:55 EDTFATAL  slon_node_health_check() returned false - fatal
health problem!

REPAIR CONFIG may be helpful to rectify this problem2010-11-04 12:10:55
EDTCONFIG slon: child terminated status: 9; pid: 29513, current worker pid:
29513
2010-11-04 12:10:55 EDTINFO   slon: done
2010-11-04 12:10:55 EDTINFO   slon: exit(0)
postgres at cbbrowne [12:11:03] [~/slony1-engine.github/src/slon] [bug156 *]
-> % cat /tmp/stderr.txt
WARNING:  table [id,nsp,name]=[1,table1,public] - sl_table does not match
pg_class/pg_namespace
WARNING:  table [id,nsp,name]=[2,table2,public] - sl_table does not match
pg_class/pg_namespace
WARNING:  table [id,nsp,name]=[4,table4,public] - sl_table does not match
pg_class/pg_namespace
WARNING:  table [id,nsp,name]=[5,table5,public] - sl_table does not match
pg_class/pg_namespace
WARNING:  Mismatch found between sl_table and pg_class.  Slonik command REPAIR
CONFIG may be useful to rectify this.
postgres at cbbrowne [12:11:07] [~/slony1-engine.github/src/slon] [bug156 *]
-> %

-- 
Configure bugmail: http://www.slony.info/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.