Is there are a reason why you can't compile your code/libraries with
-fno-omit-frame-pointer and use a fast unwinder?
Fast unwinder is nice but
* some legacy libs can't be recompiled
* people may not want to rebuild standard libraries with
-fno-omit-frame-pointer (glibc, etc.)
* I'd prefer not
In my experience, slow unwinder is so slow on ARM that it's completely
unusable. I've seen ~90% CPU time spent in the unwinder (things are
slightly better with malloc_context_size=3 or so, but what's the point
of the real unwinder then?).
Does this mean that you support fast
Should be fixed by r202845.
On Tue, Mar 4, 2014 at 4:28 PM, Alexey Samsonov samso...@google.com wrote:
I'll take a look if we can implement it easy enough.
On Tue, Mar 4, 2014 at 4:18 PM, Evgeniy Stepanov euge...@google.comwrote:
I don't mind, if it does not mess up asan code too badly.