two or more CONNECTIONTYPE WMS -> Crash

[tee]

The Mapserver is crashing down with the message

"Application Failure  mapserv.exe 0.0.0.0 in libcurl.dll 1.19.4.0 at offset 000201d9"

in the event log.

I took two public layers from this list
​http://www.lgb-rlp.de/wms-dienste.html

Here is the layer definition of both layers with CONNECTIONTYPE WMS. I tested them and the system error is prompting betimes.

LAYER
	NAME 'gewinnungsstellen_extern_wms'
	STATUS ON
	TYPE RASTER
	GROUP 'group 1'
	CONNECTIONTYPE WMS
	CONNECTION 'http://mapserver.lgb-rlp.de/cgi-bin/gewinnungsstellen?'
	METADATA
		'ows_group_title' 'group 1'
		'wms_name' 'gewinnungsstellen'
		'wms_title' 'Gewinnungsstellen mineralischer Rohstoffe'
		'wms_srs' 'epsg:31467 epsg:31466 epsg:25832 epsg:54004 epsg:4326 epsg:4258'
		'wms_style' 'default'
		'wms_server_version' '1.1.1'
		'wms_format' 'image/png'
		'wms_connectiontimeout' '60'
		'wms_style_default_legendurl_href' 'http://mapserver.lgb-rlp.de/cgi-bin/gewinnungsstellen?&SERVICE=WMS&VERSION=1.1.1&REQUEST=GetLegendGraphic&format=image/gif&LAYER=gewinnungsstellen'
		'wms_style_default_legendurl_format' 'image/gif'
		'wms_style_default_legendurl_width' '1'
		'wms_style_default_legendurl_height' '1'
	END #METADATA
END #LAYER

LAYER
	NAME 'archivfunktion_extern_wms'
	STATUS ON
	TYPE RASTER
	GROUP 'group 1'
	CONNECTIONTYPE WMS
	CONNECTION 'http://mapserver.lgb-rlp.de/cgi-bin/archivfunktion?'
	METADATA
		'ows_group_title' 'group 1'
		'wms_name' 'archivfunktion'
		'wms_title' 'Böden als Archiv der Kultur- und Naturgeschichte'
		'wms_srs' 'epsg:31467 epsg:31466 epsg:25832 epsg:54004 epsg:4326 epsg:4258'
		'wms_style' 'default'
		'wms_server_version' '1.1.1'
		'wms_format' 'image/png'
		'wms_connectiontimeout' '60'
		'wms_style_default_legendurl_href' 'http://mapserver.lgb-rlp.de/cgi-bin/archivfunktion?&SERVICE=WMS&VERSION=1.1.1&REQUEST=GetLegendGraphic&format=image/gif&LAYER=archivfunktion'
		'wms_style_default_legendurl_format' 'image/gif'
		'wms_style_default_legendurl_width' '1'
		'wms_style_default_legendurl_height' '1'
	END #METADATA
END #LAYER

The GetMap request looks like

http://localhost:80/cgi-bin
/mapserv.exe?LAYERS=gewinnungsstellen_extern_wms,archivfunktion_extern_wms&SRS=EPSG:31466&FORMAT=image/png&WIDTH=575&HEIGHT=500&SERVICE=WMS&VERSION=1.1.1&STYLES=&BBOX=2451583.925,5380340.5,2783371.575,5668851.5&map=C:/projects/umn_maps/extern_wms_demo.map&REQUEST=GetMap

See this in context ​http://osgeo-org.1803224.n2.nabble.com/two-or-more-CONNECTIONTYPE-WMS-Crash-td5478157.html#a5478157

attached you see a brakepoint that is catched be Visual Studio

Thank you very much and regards,
Daniel

[tee]

Screenshot of the catched breakpoint by Visual Studio

[dmorissette]

Alan, can you please look into this, perhaps using Valgrind on Linux will report something.

[dmorissette]

BTW, this is happening with MS4W with MapServer 5.6.3

[aboudreault]

I've tested this on Linux and everything work well. Valgrind doesn't output anything wrong.

[tee]

Thanks for your investigations.
On my machines (WinXP Prof and divers ms4w versions) the error appears not on each request, but on each fourth or tenth.

[dmorissette]

Assefa, we are unable to reproduce this on Linux, can you please see if you can reproduce on Windows with MS4W + MapServer 5.6.3?

Tee, are you sure it's not DLL Hell hitting you (i.e. multiple versions of some DLLs on the system?)

[assefa]

I will try on windows and report back.

[tee]

Unfortunately yes, I'm sure. because

1. the version in the event log is right and
   
2. I have no other libcurl.dll in the system32 folders or somewhere else and
   
3. in the cgi-bin folder is only the right libcurl.dll placed and ​http://en.wikipedia.org/wiki/DLL_hell says "... is the Windows shared library policy which favours locally (application directory) available DLLs before the one in the system library path."
   

Thanks a lot and regards, Daniel

[assefa]

I was not able to reproduce this locally using ms4w and MapServer 5.6.3. I get the expected images consistently

[aboudreault]

What should we do with this ticket? tee, Is it still valid with 5.6.5 and svn_trunk?