Gentle People:
While I thought that this was initially a libcurl problem it may be a DNS or other Fedora problem. This explains the Fedora User Group post.
I am using libcurl and some of Curl Website example code as calling routines for a multi-platform project where a webpage is read from finance.yahoo.com.
Everything works as expected on Solaris 10 and MS Windows, delivering excellent performance reading a web page.
Now when I perform the exact same access on Fedora 14 the read performance is very slow. When the read access is executed by sending the URL there seems to be a pause of one or two minutes to get a response. In the end there is always a correct response but the delay is unacceptable.
For both Solaris 10 and Fedora 14 the native gcc is used to build the project, and for Windows the Mingw Cross Environment with gcc is used. There are varying versions of gcc used. See below.
Please note that the exact same source code and make file is used in the build in all three environments.
Also the libcurl version varies in all three environments. However this morning I updated the Fedora 14 environment to the libcurl newest version 7.32 and the performance did not improve.
From Rich Grey: Could this be a server/DNS problem? I've seen long delays like this when telnetting to a server which tries to do a reverse DNS lookup on the connecting client and fails. After the lookup timeout, which can be minutes, the login: prompt finally appears and the session proceeds normally.
Please note the attached Read_Yahoo.c function that I use for the web page read.
Please note that I can post more information if requested.
- Solaris 10 gcc -dumpversion 3.4.6
- Fedora 14 gcc -dumpversion 4.5.1
Windows Mingw Cross Environment: gcc -dumpversion 4.5.1
Thomas Dineen
On 8/31/2013 11:57 AM, Thomas Dineen wrote: