Wireshark-bugs: [Wireshark-bugs] [Bug 12955] Interface list issues after system sleep/resume (du
Date: Mon, 03 Oct 2016 10:39:38 +0000

changed bug 12955


What Removed Added
CC   rknall@gmail.com

Comment # 2 on bug 12955 from
(In reply to Peter Wu from comment #1)
> After only a few (2-3) suspend cycles I noticed an increased CPU usage.
> Attaching a debugger shows 308 interfaces:
> 
> $ gdb -q -p `pidof wireshark` -ex b\ InterfaceTree::updateStatistics -ex c
> ...
> Thread 1 "wireshark" hit Breakpoint 1, InterfaceTree::updateStatistics
> (this=0x6060001003a0)
>     at /tmp/wireshark/ui/qt/interface_tree.cpp:316
> 316         if (!stat_cache_) {
> (gdb) next
> 321         if (!stat_cache_) return;
> (gdb) 
> 323         QTreeWidgetItemIterator iter(this);
> (gdb) 
> 324         while (*iter) {
> (gdb) 
> 327             for (if_idx = 0; if_idx <
> global_capture_opts.all_ifaces->len; if_idx++) {
> (gdb) p global_capture_opts.all_ifaces->len
> $1 = 308
> 
> This Wireshark binary is ASAN-instrumented and runs with
> LSAN_OPTIONS=fast_unwind_on_malloc=0 set, this slows malloc a lot which can
> be seen in the perf trace (percentages are accounting cycles):
> 98% InterfaceTree::updateStatistics
>  + 50% QString constructor
>    + 48% __sanitizer::BufferedStackTrace::SlowUnwindStack
>  + 42% ~QString
>    + 41% __sanitizer::BufferedStackTrace::SlowUnwindStack


Can you confirm, that this bug still exists with the new view/model
implementation of the main interface list, or can be close it?


You are receiving this mail because:
  • You are watching all bug changes.