Version: (using KDE KDE 3.2.3) Installed from: Gentoo Packages Compiler: gcc version 3.3.4 20040623 (Gentoo Linux 3.3.4-r1, ssp-3.3.2-2, pie-8.7.6) OS: Linux I frequently see kio_http going to high cpu utilization: kdeinit: kio_http https /tmp/ksocket-carl/klauncheryFqgAa.slave-socket /tmp/ksocket-carl/konquerordMaJhc.slave-socket When this is happening, there will be a tab in konq that is stuck loading all or part of an https page. Closing the tab or window, or reloading the page will return things to normal. Generally this condition is associated with more than one page trying to load https pages or elements. For example, I frequently have open: https://orbit.theplanet.com/index.html Which refreshes itself periodically to show a network utilization graph. Normally, this runs just fine, although when I start working intensively in another https page in another tab, this bug presents itself. I have been seeing this for quite awhile, like maybe as far back as 3.1, but have never seen any other bug reports on it. Sorry to take so long to report it. Also, there is a proxy involved (Squid).
Can you attach strace to the http process when it is doing this? You can use: strace -tt -p <pid of process> Do you have kdelibs compiled with debug info? If so, you may also want to have a look with gdb what is going on: gdb kdeinit <pid of process> and once it is finished loading, type: bt to get a backtrace.
Sorry to take so long with this. Ater writing the bug, I tried turning of the proxy, and didn't see the problem until just now, after I turned it back on. Here's what ptrace shows: # strace -tt -p 31793 Process 31793 attached - interrupt to quit 11:39:31.870910 read(10, "", 5) = 0 11:39:31.871117 getpid() = 31793 11:39:31.871200 read(10, "", 5) = 0 11:39:31.871274 getpid() = 31793 11:39:31.871339 read(10, "", 5) = 0 11:39:31.871396 getpid() = 31793 11:39:31.871467 read(10, "", 5) = 0 11:39:31.871531 getpid() = 31793 11:39:31.871597 read(10, "", 5) = 0 11:39:31.871658 getpid() = 31793 11:39:31.871721 read(10, "", 5) = 0 11:39:31.871779 getpid() = 31793 11:39:31.871845 read(10, "", 5) = 0 11:39:31.871901 getpid() = 31793 11:39:31.871969 read(10, "", 5) = 0 11:39:31.872025 getpid() = 31793 11:39:31.872095 read(10, "", 5) = 0 11:39:31.872158 getpid() = 31793 11:39:31.872225 read(10, "", 5) = 0 11:39:31.872283 getpid() = 31793 11:39:31.872347 read(10, "", 5) = 0 11:39:31.872406 getpid() = 31793 11:39:31.872508 read(10, "", 5) = 0 11:39:31.872571 getpid() = 31793 11:39:31.872633 read(10, "", 5) = 0 11:39:31.872693 getpid() = 31793 11:39:31.872756 read(10, "", 5) = 0 11:39:31.872817 getpid() = 31793 11:39:31.872874 read(10, "", 5) = 0 I don't have kdelibs compiled with debug symbols, so I can't there, sorry.. FWIW, I have all three proxy server settings (http, https, and ftp) pointing to a local Squid server: http://proxy 3128 hth
I'm experiencing the same problem. I'm using a monitoring site daily via https, and it does seem that the https sessions are the ones that lock up. KDE 3.2, Mandrake 10 I rebuilt kdelibs with --debug=full and ran a gdb backtrace. This morning I had two kio_http processes locked, both of them competing for 100% CPU utilization. Backtrace of the first one: #0 0x41767616 in BIO_new_socket () from /usr/lib/libcrypto.so.0.9.7 The second one displayed this consistantly: #0 0x4021a580 in QMemArray<char>::resize(unsigned) () from /usr/lib/libkio.so.4 Until I killed the first one. The second one continued to chew 100% CPU and backtrace now says: #1 0x00000010 in ?? () #2 0x080b6d48 in ?? () #3 0x00000010 in ?? () #4 0x40ecbd3c in QString::truncate(unsigned) () from /usr/lib/qt3/lib/libqt-mt.so.3 #5 0x00000010 in ?? () #6 0x4103e998 in ?? () from /usr/lib/qt3/lib/libqt-mt.so.3 #7 0xffffffff in ?? () and a short while later (no noticible change) says: #0 0x40756a50 in kdbgstream::flush() () from /usr/lib/libkdecore.so.4 #1 0xbfffc380 in ?? () #2 0x4104e5f4 in QString::shared_null () from /usr/lib/qt3/lib/libqt-mt.so.3 #3 0x080a21a8 in ?? () #4 0x080515f1 in QCharRef::operator QChar() const () Hope this helps. I am also using a Squid proxy. Let me know if I can be of further assistance - this bug is annoying me to no end, I run into it a dozen or more times daily at the moment.
No one else running into this? It's driving me nuts.
FWIW, one workaround is to not use the proxy for https (http only seems to work OK).
Sadly, firewall rules necessitate https via proxy. Is this bug evident in the latest KDE? I'm still using 3.2.0.
Has anyone here tried to see if the problem exists in the newer KDE releases (3.3.1) ? I personally do not see such issues with the current development branch.
I still see this problem with https that redirects to http (for example, secure login to "my ebay" or amazon). kde version is 3.4 here on suse linux.
disabling proxy usage makes https work. Sadly, going around the proxy is a "no-no" for me.
Too old... Feel free to reopen if problem persists in current version of KDE.