[Mapserver-dev] connection pooling segfaults SDE

[Jan Hartmann]

Howard Butler wrote:

> There still are some bugs with connection pooling that cause my stuff to 
> segfault from time to time.  After I upgraded to SDE 9.0, the strategy 
> that I took to prevent them didn't work anymore.  I will be getting back 
> to this problem soon, but I don't think it is an issue here.
> 

Howard,

I submitted  the global connection pooling mechanism at the beginning of 
2003, and Steve Lime then adapted it for SDE and committed it. There 
seemd to have been no problems since then until you accessed it with 
Python MapScript. The problem could have to do with two points I 
discussed with Steve at that time. Perhaps they can help you in the 
right direction.

Jan

---------------------------------------------------------------------------
(email 22-01-2003)

Jan Hartmann wrote:
 > Steve Lime wrote:
 >
 >> I've commited the intitial version of the pooling for SDE. Seems to
 >> work  just fine and less bookkeeping issues with this implentation
 >> than the first so should be friendlier with MapScript and so on...
 >>
 >
 > Great! If all problems had such simple solutions... I set up
 > everything for PostGIS (it really has become very simple!), and sent
 > the patches to Dave Blasby. BTW: no problems with PHP Mapscript here.
 >

 > Two last remarks:
 >
 > - When using the pooled CVS version with unaltered PostGIS support,
 > each PostGIS layer gets closed twice: in msLayerClose() and in
 > msCloseConnections(). This sometimes lets MapServer wait for ever
 > (don't know exactly when this happens). Anyway, as long as pooling is
 > not enabled for PostGIS, the call to msPOSTGISLayerCLose() should be
 > commented out in msCloseConnections()
 >
 > - msCheckConnection() does indeed assume (as you comment) that the
 > order in which layers are created is fixed. If layers are shuffled
 > around before all layers are created the algorithm doesn't work. This
 > is only a problem if new layers are created after reordering, not if
 > reordering takes place after all layers have already been created. I
 > don't think this is a problem with MS CGI, but I can imagine that some
 > scripts could intermingle layer creation and shuffling.