8.0.0rc1 on hppa2.0w-hp-hpux11.00

Started by Christoph Hallerover 21 years ago5 messagesbugs
Jump to latest
#1Christoph Haller
ch@rodos.fzk.de

gmake check result

template1=# select version();
version
-------------------------------------------------------------------------------
PostgreSQL 8.0.0rc1 on hppa2.0w-hp-hpux11.00, compiled by GCC gcc (GCC)
3.3.1
(1 row)

1 of 96 tests failed

test stats ... FAILED

Please find attached

src/test/regress/regression.out
src/test/regress/regression.diffs

BTW, as mentioned the other day I had to do configure as

configure LIBS="-ltermcap" ...

From my previous mail

I have successfully compiled and installed 7.4.5 yesterday.
After doing 'make' I saw the line "Ready to install".
template1=# select version();
version
--------------------------------------------------------
PostgreSQL 7.4.5 on hppa2.0w-hp-hpux11.00, compiled by GCC gcc (GCC)
3.3.1
(1 row)

But when I started psql it immediately core dumped with
libreadline.sl.5 Unresolved external: tgetent

I did a search on the mailing list archives and got 63 hits on
'tgetent',
dating from 1997 to 2004, all having the same problem,
the missing '-ltermcap' in LIBS in src/Makefile.global.

After distclean and doing configure LIBS="-ltermcap" ...
and make again it finally worked.

In http://archives.postgresql.org/pgsql-bugs/2001-09/msg00023.php
Peter Eisentraut writes

LIBS = -lz -lresolv -lcompat -lm -lutil -ltermcap -lreadline

That's a good fix.

The current CVS tip also works around this problem.

But the 7.4.5 src/Makefile.global still has this:
LIBS = -lz -lreadline -lPW -lgen -lBSD -ldld -lnsl -lm

So it appears this fix must have been lost anywhere.

And it is not fixed in 8.0.0rc1.
8.0.0rc1 src/Makefile.global still has this:
LIBS = -lz -lreadline -lPW -lgen -lBSD -ldld -lnsl -lm

Regards, Christoph

Attachments:

regression.diffstext/plain; charset=iso-8859-2; name=regression.diffsDownload+6-6
regression.outtext/plain; charset=iso-8859-2; name=regression.outDownload
#2Tom Lane
tgl@sss.pgh.pa.us
In reply to: Christoph Haller (#1)
Re: 8.0.0rc1 on hppa2.0w-hp-hpux11.00

Christoph Haller <ch@rodos.fzk.de> writes:

test stats ... FAILED

Check the postmaster.log file to find out why the stats collector didn't
start.

regards, tom lane

#3Christoph Haller
ch@rodos.fzk.de
In reply to: Christoph Haller (#1)
Re: 8.0.0rc1 on hppa2.0w-hp-hpux11.00

In ./src/test/regress/log/postmaster.log I cannot find
anything regarding the stats collector non-start.
Does it help to send the file (about 72 KB)?

Regards, Christoph

Tom Lane wrote:

Show quoted text

Christoph Haller <ch@rodos.fzk.de> writes:

test stats ... FAILED

Check the postmaster.log file to find out why the stats collector didn't
start.

regards, tom lane

#4Tom Lane
tgl@sss.pgh.pa.us
In reply to: Christoph Haller (#3)
Re: 8.0.0rc1 on hppa2.0w-hp-hpux11.00

Christoph Haller <ch@rodos.fzk.de> writes:

In ./src/test/regress/log/postmaster.log I cannot find
anything regarding the stats collector non-start.

Hmm, I'd expect there to be a complaint right at the beginning.

Does it help to send the file (about 72 KB)?

Off-list please.

regards, tom lane

#5Tom Lane
tgl@sss.pgh.pa.us
In reply to: Christoph Haller (#1)
Re: 8.0.0rc1 on hppa2.0w-hp-hpux11.00

Christoph Haller <ch@rodos.fzk.de> writes:

In ./src/test/regress/log/postmaster.log I cannot find
anything regarding the stats collector non-start.

LOG: could not translate host name "localhost", service "65432" to address: Unknown host
WARNING: could not create listen socket for "localhost"
LOG: could not resolve "localhost": Unknown host
LOG: database system was shut down at 2004-12-15 13:56:31 MET
...

I believe the first two messages relate to being unable to open a TCP/IP
postmaster socket, which is not fatal because we can still listen on a
Unix socket; and the third one indicates that we failed to open a UDP
socket for the stats collector. In any case you've got to fix your name
resolution setup.

Memo to list: I'm inclined to reshuffle the code in pgstat_init() so
that the notice "disabling statistics collector for lack of working
socket" is emitted in *all* cases where we fail to start the stats
collector, rather than only some cases as now. Any objections?

regards, tom lane