amule~2.3.1-r2 works fine for lots of weeks, but after upgrading to 2.3.2 it eats up all my memory: top - 16:52:22 up 1:55, 1 user, load average: 9,13, 6,08, 2,92 Tasks: 342 total, 3 running, 338 sleeping, 0 stopped, 1 zombie %CPU(s): 3,0 us, 2,2 sy, 0,0 ni, 57,2 id, 37,5 wa, 0,0 hi, 0,1 si, 0,0 st KiB Spch: 12367508 total, 119004 free, 12036820 used, 211684 buff/cache KiB Swap: 16383996 total, 578352 free, 15805644 used. 131872 avail Spch PID USER PR NI VIRT RES SHR S %CPU %MEM ZEIT+ BEFEHL 27040 andy 20 0 25,628g 0,011t 4836 R 98,7 91,3 8:02.55 amule I'm not the only one: http://bugs.amule.org/view.php?id=1770 duplicate of this bug: http://bugs.amule.org/view.php?id=1768 Therefore the version 2.3.2 should better be masked, as I did it by myself.
Same here, my perf top looks like this: PerfTop: 5253 irqs/sec kernel:65.6% exact: 100.0% [4000Hz cycles:ppp], (all, 8 CPUs) ------------------------------------------------------------------------------------------------------------------------- 11.80% [kernel] [k] mwait_idle_with_hints.constprop.3 6.29% libncursesw.so.6.0 [.] _nc_render 2.30% libglib-2.0.so.0.4800.2 [.] g_mutex_lock 2.17% libglib-2.0.so.0.4800.2 [.] g_mutex_unlock 1.98% libncursesw.so.6.0 [.] wadd_wchnstr 1.85% [kernel] [k] do_sys_poll 1.69% [kernel] [k] copy_user_enhanced_fast_string 1.41% [kernel] [k] __fget 1.17% [kernel] [k] do_raw_spin_lock 1.10% [kernel] [k] syscall_return_via_sysret 1.10% [kernel] [k] copy_msghdr_from_user 1.09% [kernel] [k] audit_filter_syscall 1.07% [kernel] [k] atomic64_dec_and_test 1.05% [kernel] [k] rcu_eqs_enter_common 0.98% libncursesw.so.6.0 [.] whline 0.92% libc-2.23.so [.] wcwidth 0.90% [kernel] [k] entry_SYSCALL_64 0.88% libglib-2.0.so.0.4800.2 [.] g_main_context_check 0.86% libncursesw.so.6.0 [.] wnoutrefresh 0.81% libc-2.23.so [.] 0x0000000000147b68 0.80% [kernel] [k] __audit_syscall_exit 0.77% [kernel] [k] native_write_msr 0.76% [kernel] [k] _raw_spin_lock_irqsave 0.75% [kernel] [k] __rcu_read_unlock 0.72% [kernel] [k] genregs_get 0.71% [kernel] [k] rcu_eqs_exit_common.isra.32 0.70% htop [.] RichString_append 0.67% [kernel] [k] __audit_syscall_entry 0.66% [kernel] [k] __rcu_read_lock 0.65% [kernel] [k] syscall_trace_enter 0.64% libglib-2.0.so.0.4800.2 [.] g_main_context_prepare 0.57% libncursesw.so.6.0 [.] _nc_hash_map_sp 0.56% libpthread-2.23.so [.] pthread_mutex_lock 0.55% libpthread-2.23.so [.] __pthread_mutex_unlock_usercnt 0.54% [kernel] [k] audit_filter_rules 0.53% libc-2.23.so [.] vfprintf 0.50% [kernel] [k] do_nocb_deferred_wakeup
Could you try to report it to github too? It seems they are more active there :/ https://github.com/amule-project/amule/issues
What if WX_GTK_VER="2.8" is set in the ebuild file instead of the version 3.0?
(In reply to David Kredba from comment #3) > What if WX_GTK_VER="2.8" is set in the ebuild file instead of the version > 3.0? Tested today with version 2.3.2-r1 and that fixes the problem for me. Before that, I tried amule-2.3.2-r1 with the unmodified ebuild and the same memory leak happened. Many thanks for this hint, Andy.
Same problem here. aMule has just given me: (amule:12900): GLib-CRITICAL **: Source ID 2946 was not found when attempting to remove it
If you can add a comment in upstream report, please tell them that this is an wxGTK3 related issue thanks
Or, maybe, it should be reported here instead : https://github.com/amule-project/amule/issues
I put more information on http://bugs.amule.org/view.php?id=1770. Let's see what will happen. Meanwhile, I tried #c3 (WX_GTK_VER="2.8"). It works fine. Maybe we can just switch wxGTK version.
We will need to treeclean this if continues to not work with wxGTK3
Or fallback to WX_GTK_VER="2.8"
It seems the solution is to switch from wxSocketEvent to Boost.Asio (--with-boost), and it is preferred by upstream. I will fix this issue in github PR and will take maintainership of this package if nobody disagree.
Sure, you are welcome
(In reply to Alexander Tsoy from comment #11) > It seems the solution is to switch from wxSocketEvent to Boost.Asio > (--with-boost), and it is preferred by upstream. I will fix this issue in > github PR and will take maintainership of this package if nobody disagree. Will you be pushing this to the main tree or an overlay? Id like to be updated on its progress. I currently use the ebuild from "stefantalpalaru" overlay because the one in the main tree doesnt build. But it has the same issue as is referenced in this bug. It required a call to "setup-wxwidgets" as required by "wxwidgets.eclass". as referenced in https://bugs.gentoo.org/654692 that issue is stated to have already been fixed upstream though.
(In reply to jeremy mills from comment #13) > Will you be pushing this to the main tree or an overlay? Id like to be > updated on its progress. To the main tree of course. You can follow PR on github. > It required a call to "setup-wxwidgets" as required by "wxwidgets.eclass". > as referenced in https://bugs.gentoo.org/654692 that issue is stated to > have already been fixed upstream though. The ebuild currently calls need-wxwidgets() which is a wrapper around setup-wxwidgets(), so this change is mostly cosmetic. The real issue is an API mismatch with crypto++-6 which is fixed upstream.
The bug has been closed via the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=4ca0547326aae8ebba52da90a6c2199eeffcf790 commit 4ca0547326aae8ebba52da90a6c2199eeffcf790 Author: Alexander Tsoy <alexander@tsoy.me> AuthorDate: 2018-06-18 17:52:42 +0000 Commit: Patrice Clement <monsieurp@gentoo.org> CommitDate: 2018-06-22 22:17:13 +0000 net-p2p/amule: build with dev-libs/boost. Switch from wxSocketEvent to Boost.Asio to prevent memory leak and complete lockup. Closes: https://bugs.gentoo.org/597194 Closes: https://github.com/gentoo/gentoo/pull/8892 net-p2p/amule/amule-2.3.2-r2.ebuild | 105 ++++++++++++++++++++++++++++++++++++ 1 file changed, 105 insertions(+)
