RE: 100% CPU usage ?
"Deccio, Casey T" <[email protected]>
| Newsgroups | gmane.network.nocat |
|---|---|
| Message-ID | <[email protected]> |
I'm not sure if this is the same bug that I had been experiencing earlier, but I had fixed something similar and submitted a patch (though it didn't really come through on the mail list, so I think I mailed it to Rob)--my patch involved upgrading to glib2 because there seemed to be issues with the main loop watching the fd to see the second packet (second part of a request) coming through with glib1. This after I submitted an earlier patch providing a blocking solution because I couldn't get glib1 to work. I'm not a glib expert, and I'm not sure if glib2 is significantly bigger than glib1 and that's why it wasn't being used for NoCatSplash. Anyway, I don't have access to the code now, but I did some testing with handle_request and http_request_read, so you may see if it's along the lines of the patch you're fixing. Casey -----Original Message----- From: [email protected] on behalf of Schuyler Erle Sent: Sat 5/14/2005 11:45 AM To: Simon Dimenstein Cc: [email protected] Subject: Re: [NoCat] 100% CPU usage ? * On 13-May-2005 at 1:24PM PDT, Simon Dimenstein said: > I have the same issue. > Here is snipet from strace. > It dies when it goes in to read loop Thanks for this report, Simon. Based on this observation, I've checked an updated http.c into CVS that checks the number of bytes read by g_io_channel_read() and bails out of the loop if the answer is zero. I feel a bit silly for not adding this check earlier. It compiles all right on my system, but I don't have the means to test the gateway properly right this moment. Can anyone download and build and test the latest version from SourceForge CVS and give me some idea of whether or not it works? http://sourceforge.net/cvs/?group_id=118349 SDE