Nilsson Bengt wrote:
Hi,
I'm not sure that it's a fastcgi problem.
As you can see from the postgresql log, postgresql reports a
warning: 'WARNING: 25001: there is already a transaction in progress'.
The cause of that, I guess, is that mapserver, in mappostgis.c, sends
(another) 'BEGIN'. During the first call it also sent a BEGIN. In
mappostgis.c
there is a check on the outcome of the BEGIN (if(!result ||
PQresultStatus
(result) != PGRES_COMMAND_OK) { ) and I guess that it won't be OK
because
of the warning.
Digging around in the postgresql.conf file gave me an idea of changing
the '#client_min_messages = notice' to 'client_min_messages = error'. As
I
understand that means that postgresql won't report warnings back to the
client, mappostgis, and thus the test won't fail.
I haven't got any 'Segmentation faults' from the application when
running it as a static server since my change so I guess I'm on the
right track.
I wonder why mapserver sends repeating BEGINs and no ENDs. I
can't find any code for sending an END. Normally, no fastcgi with
persistant
connections, it won't matter since I guess that a closed connection also
means 'END'.
Bengt,
Good analysis!
This should certainly be filed as a bug report and I'd appreciate
being added in the cc: list. Likely it should be primarily assigned
to whoever owns the postgis component.
I classify this as loosely a fastcgi problem since it is a problem
that would only come out in fastcgi use (or perhaps long running
mapscript applications).
I *did* use fastcgi with postgis a few years ago so either things
have regressed or there is some specific aspect of your configuration
that tickles the problem.
Best regards,
--
---------------------------------------+--------------------------------------
I set the clouds in motion - turn up | Frank Warmerdam, [EMAIL PROTECTED]
light and sound - activate the windows | http://pobox.com/~warmerdam
and watch the world go round - Rush | President OSGeo, http://osgeo.org
