[mapserver-users] MapCache: Disabling WMS requests for fully-seeded caches?

[thomas bonfort]

I spoke a bit too soon, the segfault is fixed, but the empty image is
returned with a text/plain mimetype which is incorrect.
I don't have the time to look into this further now, could you please
open a ticket for this so the issue isn't forgotten?

tks,
thomas

On Tue, Nov 29, 2011 at 22:27, thomas bonfort <thomas.bonfort at gmail.com> wrote:
> hmm, ok, you're using the mixed image format for the empty image, and
> that was not a supported use-case. I've just commit a fix that fixes
> the issue for me here.
>
> regards,
> thomas
>
> On Tue, Nov 29, 2011 at 22:13, thomas bonfort <thomas.bonfort at gmail.com> wrote:
>> I'll have a look, but my config that doesn't segfault is using JPEG.
>> Note that the 404 error code is the way it's supposed to work, if you
>> inspect the headers and body you should see a image/jpeg mimetype, a
>> jpeg body, and an X-MapCache-Error header explaining what happened.
>>
>> --
>> thomas
>>
>> On Tue, Nov 29, 2011 at 21:22, John Taranu
>> <jtaranu at firstbasesolutions.com> wrote:
>>> Note that setting the default format to PNG doesn't cause a segmentation
>>> fault:
>>>
>>>
>>>   <default_format>PNG</default_format>
>>>   <errors>empty_img</errors>
>>>
>>> However, in this case mapcache still returns a HTTP 404 instead of a blank
>>> image.
>>>
>>>
>>>
>>> -----Original Message-----
>>> From: Travis Kirstine [mailto:traviskirstine at gmail.com]
>>> Sent: Tuesday, November 29, 2011 3:21 PM
>>> To: thomas bonfort
>>> Cc: John Taranu; mapserver-users at lists.osgeo.org
>>> Subject: Re: [mapserver-users] MapCache: Disabling WMS requests for
>>> fully-seeded caches?
>>>
>>> Thomas,
>>>
>>> Here is the backtrace:
>>>
>>> (gdb) run -X
>>> Starting program: /usr/sbin/httpd -X
>>> warning: no loadable sections found in added symbol-file system-supplied DSO
>>> at 0x2aaaaad1b000 [Thread debugging using libthread_db enabled]
>>>
>>> Program received signal SIGSEGV, Segmentation fault.
>>> 0x00002aaaaca4e32b in strstr () from /lib64/libc.so.6
>>> (gdb) bt
>>> #0  0x00002aaaaca4e32b in strstr () from /lib64/libc.so.6
>>> #1  0x00002aaab539737d in mapcache_image_create_empty () from
>>> /usr/lib64/httpd/modules/mod_mapcache.so
>>> #2  0x00002aaab53971fe in mapcache_configuration_parse_xml () from
>>> /usr/lib64/httpd/modules/mod_mapcache.so
>>> #3  0x00002aaab53a047a in mapcache_configuration_parse () from
>>> /usr/lib64/httpd/modules/mod_mapcache.so
>>> #4  0x00002aaab53879e6 in mapcache_add_alias (cmd=0x7fffffffe760, cfg=<value
>>> optimized out>, alias=0x2aaaaae0f850 "/mapcache",
>>>    configfile=0x2aaaaae0f860 "/var/www/html/mapcache/mapcache.xml")
>>> at mod_mapcache.c:502
>>> #5  0x00002aaaaaad5c4b in ?? ()
>>> #6  0x00002aaaaaad60d0 in ap_walk_config ()
>>> #7  0x00002aaaaaad6232 in ap_process_config_tree ()
>>> #8  0x00002aaaaaac1c51 in main ()
>>> (gdb)
>>>
>>> #################################################################
>>>
>>> MapCache snippet:
>>>
>>>  <default_format>mixed</default_format>
>>>
>>>  <errors>empty_img</errors>
>>>
>>> Regards
>>>
>>>
>>>
>>>
>>> On 29 November 2011 11:56, thomas bonfort <thomas.bonfort at gmail.com> wrote:
>>>> I've just tried report_img locally without issues, any chance you can
>>>> provide a backtrace ?
>>>>
>>>> On Tue, Nov 29, 2011 at 17:32, John Taranu
>>>> <jtaranu at firstbasesolutions.com> wrote:
>>>>> Hi Thomas,
>>>>>
>>>>> Thanks for the suggestion.  Commenting out the source block disables the
>>> WMS requests.
>>>>>
>>>>> However, the error reporting seems to be broken.  Setting
>>> <errors>empty_img</errors> crashes Apache with a segmentation fault
>>> (mod_mapcache SVN revision 12828).  Setting <errors>report</errors> throws
>>> HTTP 404 responses, which is a more reasonable result.
>>>>>
>>>>> John
>>>>>
>>>>>
>>>>>
>>>>> -----Original Message-----
>>>>> From: thomas bonfort [mailto:thomas.bonfort at gmail.com]
>>>>> Sent: Tuesday, November 29, 2011 10:33 AM
>>>>> To: John Taranu
>>>>> Cc: mapserver-users at lists.osgeo.org
>>>>> Subject: Re: [mapserver-users] MapCache: Disabling WMS requests for
>>> fully-seeded caches?
>>>>>
>>>>> You can do this by commenting out the <source> block from the tileset
>>> once the wanted tiles have been created. You'll probably also want set error
>>> reporting to return blank images from that stage onwards.
>>>>>
>>>>> --
>>>>> thomas
>>>>>
>>>>> On Tue, Nov 29, 2011 at 00:09, John Taranu
>>> <jtaranu at firstbasesolutions.com> wrote:
>>>>>> We have a fully seeded cache that was fully seeded to a polygon outline.
>>>>>> The intent is to serve the tileset through a variety of services
>>>>>> (WMS,
>>>>>> TMS) using mod-mapcache.
>>>>>>
>>>>>>
>>>>>>
>>>>>> Tiles that are outside the polygon outline but still within the
>>>>>> polygon bounding box should be blank tiles.  However, a request for
>>>>>> one of these tiles results in a call to that tileset’s WMS source.
>>>>>> This creates considerable overhead by starting up MapServer and
>>>>>> writing the resulting tile to the cache.  Is there any way to
>>>>>> disable the WMS calls for these blank tiles, and instead
>>>>>> short-circuit the request to return a blank?  That is, once a
>>>>>> tileset has been fully seeded, is it possible to disable any further
>>>>>> calls to the WMS, so that requests for tiles that are not in the cache
>>> return a blank?
>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> mapserver-users mailing list
>>>>>> mapserver-users at lists.osgeo.org
>>>>>> http://lists.osgeo.org/mailman/listinfo/mapserver-users
>>>>>>
>>>>>
>>>> _______________________________________________
>>>> mapserver-users mailing list
>>>> mapserver-users at lists.osgeo.org
>>>> http://lists.osgeo.org/mailman/listinfo/mapserver-users
>>>
>>>
>>
>