Re: any libcurl API to delete expired cookies?
On Fri, 28 Jun 2013, YAMADA Yasuharu wrote: I think the feature of removing expired cookies should be separate from the one that limits the total amount of cookies. I'm convinced there are people who might want that expiring feature without the max limit set. I quarry expired cookies patch from my [new feature] Maximum limitation of cookies patch. Please check. Hello again Yamada, I found this old patch by accident lingering around and I thought I'd take a look and see if we can merge this. I can't apply it since both git and patch says the patch file is illegally formatted. Could you please give it a quick check and if possible submit a fixed version that I can try out? -- / daniel.haxx.se --- List admin: http://cool.haxx.se/list/listinfo/curl-library Etiquette: http://curl.haxx.se/mail/etiquette.html
Re: Curl build on HP-UX failed
On Tue, 10 Sep 2013, Arunav Sanyal wrote: It just gives a warning. I think(based on what I understand from the error), the compiler is supposed to return an error and the compilation should fail. Is my understanding correct? If yes, then I have to find the appropriate compiler flag for the same. Yes, that's the purpose of the test according to m4/curl-compilers.m4 where it is located but I'm not sure why we need to fail the build because of this. That's why I asked Yang about this since I believe he added that check so he would be a good person to elaborate on its purpose. I'm thinking it may be far from easy to find such a compiler flag, or possibly even impossible. You can easily just comment out CURL_CHECK_COMPILER_PROTOTYPE_MISMATCH from configure.ac, re-run autoconf and then run configure again and that failure shouldn't bother you. -- / daniel.haxx.se --- List admin: http://cool.haxx.se/list/listinfo/curl-library Etiquette: http://curl.haxx.se/mail/etiquette.html
Re: Make DNS requests follow the CURLOPT_INTERFACE if c-ares is in use
On Sat, 7 Sep 2013, Kim Vandry wrote: I have addressed both of these issues by making the new options take strings instead of binary IP addresses. Please find two patches at https://github.com/vandry/curl , one for libcurl and one for the command line tool. Nice work. These looks really nice, small and clean now and I'm about to merge them. Have you the curl command line options documented as well? I figure the docs can be pretty much copy and pasted from the libcurl options. -- / daniel.haxx.se --- List admin: http://cool.haxx.se/list/listinfo/curl-library Etiquette: http://curl.haxx.se/mail/etiquette.html
What happened to this old issue - Libcurl stops when sending lots of https messages
Hi everyone! We incorporated a suggested patch in our product to handle the case of hangs occurring when running heavy traffic over HTTPS using libcurl. The original issue was posted here (actually this link is the last reply in the thread and contains the fix proposal): http://curl.haxx.se/mail/lib-2009-11/0191.html The issue was caused by the fact that OpenSSL returns SSL_ERROR_WANT_WRITE on a call to SSL_read, but libcurl then waits on select(), sees that a sockets becomes writable and calls SSL_write, as opposed to calling SSL_read again as the documentation for OpenSSL suggests. The suggested fix takes this into account and correctly forces a call to SSL_read. However, this was in a rather old libcurl release 7.19.7. I've tried to reproduce the error on the latest libcurl but I'm having trouble doing so. None the less, comparing the code for Curl_readwrite in transfer.c, the function seems not to have changed all that much. So what I'm asking is if libcurl handles SSL_ERROR_WANT_READ/WRITE by calling the same OpenSSL method that caused the error return or is the strategy still as in the old version; libcurl makes a call based on whatever select() returns? Thank you, Mehmet. --- List admin: http://cool.haxx.se/list/listinfo/curl-library Etiquette: http://curl.haxx.se/mail/etiquette.html