Jonathan Houser wrote:
Aarno,
IMHO, going that far back is counterproductive. But if Kannel leaks on error,
failed fetchs should generate leaks, do they not ? And then we will have a
reproducable bug.
I was just comparing the oldest version I tested that still leaked against mine that doesn't. I did the diff of those two to minimize the size of the diff file.
I tried to break it with a simulator asking for URLs like "http://catchow", "http://catchow.ruffy.com/" and "http://10.10.0.3" with no luck. I then tried to log in to Hotmail...
http://mobile.msn.com/ select 'MSN Hotmail'
...with my simulator (Ericsson WapIDE 3.2.1), which uses OpenSSL, and that worked. It failed like the previous OpenSSL lines I've pasted and inc'ed the CLOSE_WAIT count. Below are the lines that *could* apply to it. (As an aside, is there any way to log an IP address with each line? It'd *really* help in debugging. Then I could pull lines that apply to my IP vs. some carrier's IP in cases like this.)
ok, so the assumption is: We get leaking file descriptors when HTTPS addresses are requested and some openssl write error occures, right?
This may be also reproducable, right?
Ok, but this does not clear-up the fact that your 2004-11-05 version seems to "not" leak any fds?
Stipe
mailto:stolj_{at}_wapme.de ------------------------------------------------------------------- Wapme Systems AG
Vogelsanger Weg 80 40470 Düsseldorf, NRW, Germany
phone: +49.211.74845.0 fax: +49.211.74845.299
mailto:info_{at}_wapme-systems.de http://www.wapme-systems.de/ -------------------------------------------------------------------