I have run into strange behavior (100% cpu usage) that went away when filtering out the CFLAGS Reproducible: Always Steps to Reproduce: 1. emerge net-p2p/mldonkey-2.6.7 while having "-fomit-frame-pointer" in your CFLAGS 2. /etc/init/mldonkey start Actual Results: eats all idle cpu time Expected Results: shouldn't I don't know how to do ebuilds, so no patch here. but I figure there must be some easy way to filter out unwanted cflags from /etc/make.conf on a per-package basis. can anyone please do so for the ebuild?
Could you provide your emerge info?
Its a known bug, Cryptopp C++ code does not like this flag. Current CVS already tries to filter it out by itself: 2005/11/05 4589: CryptoPP doesn't like -fomit-frame-pointer, so remove it (przemoc) Please test, until 2.7.0 is released its advised to update to ebuild so Gentoo users avoid the problem.
http://savannah.nongnu.org/patch/?func=detailitem&item_id=4589#attached Do we wait for upstream to release a version containing this patch?
(In reply to comment #3) > http://savannah.nongnu.org/patch/?func=detailitem&item_id=4589#attached > Do we wait for upstream to release a version containing this patch? I would like to hear if this patch/or not using that CFLAG solves the problem before releasing a new version. 2.7.0 should work without problems on Gentoo. A user on the german forum already reported success.
Filtering CFLAGS should take place in MLDonkey ebuild, notes from Ocaml ebuild: * Building ocaml with unsafe CFLAGS can have unexpected results * Please retry building with safer CFLAGS before reporting bugs * Likewise, building with a hardened gcc is not possible. Also hardened GCC should be handled by the ebuild. MLdonkey 2.7.0 needs lablgtk-2.6.0 (not in portage, see bug #107501), because of this patch: 2005/10/31 4583: GTK2 mlgui improvement (su_blanc) Lablgtk-2.6.0 is needed for GTK2 mlgui, it fixes a mem leak with GdkPixbuf.
RE #4: not using that cflag solves the problem for me. I have not tried the patch, but it looks like it should do the same trick.
MLDonkey compiled with -fomit-frame-pointer also creates a wrong EDK_client_private_key: client_private_key = "/root/.mldonkey" in my example. Normally it should start with "MIIBC", over 350 chars in total. I will add a check in MLDonkey to change the key if it is wrong.
Please check this patch: https://savannah.nongnu.org/patch/index.php?func=detailitem&item_id=4631
MLDonkey 2.7.0 has been released
i've added filtering -fomit-frame-pointer in 2.6.7. sorry it took so long, but i'm kind of away all the time;/ i'll try to test and submit into cvs 2.7.0 tommorow, but i can't promise i'll be able to do it