Le 26 septembre 2010, Jon Burgess a écrit : > On Sat, 2010-09-25 at 23:30 +0200, Guilhem Bonnefille wrote:
> What I think happens is: > 1) We first do a request for a tile with an ETag and apply a custom > header, this gets set into the conn->data->set.headers pointer > 2) The header gets freed, but the set.headers pointer is left as a > dangling reference to the memory > 3) A subsequent request is generated for a tile without an etag so we > do not overwrite the set.headers pointer and it keeps the old, invalid > value and the HTTP request code tries to reference it. Yes, that makes sense, and that explains Guilhem's failure. I think that I never hit the bug because I'm using only Mapnik tiles, which uses etag, and not any other tile server that doesn't use etag. Guilhem: to make things clearer, curl uses the option CURLOPT_HTTPHEADER to add any header, and expect a valid string there. Viking also reuse curl from the previous call of curl_download(), which means that CURLOPT_HTTPHEADER wasn't cleared. Btw, shouldn't we initialize _all_ curl options at each pass ? I think we can the same problem with other curl options that are initialized each time - like CURLOPT_TIMECONDITION or CURLOPT_TIMEVALUE. I will try to provide a patch to initialize everything. Thank you both for having detected and fixed the problem :) Jocelyn ------------------------------------------------------------------------------ Start uncovering the many advantages of virtual appliances and start using them to simplify application deployment and accelerate your shift to cloud computing. http://p.sf.net/sfu/novell-sfdev2dev _______________________________________________ Viking-devel mailing list Viking-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/viking-devel Viking home page: http://viking.sf.net/
