Enabling profiling causes hang?
brock.pytlik at gmail.com
Wed Feb 11 17:00:13 PST 2015
Well ... when I do objdump --dwarf on ardb-server, I get a bunch of info.
When I run ardb under gdb, I'm able to stop and get backtraces w/ symbolic
info. When I use jemalloc 3.6.0 (on x86), I get useful info. Given all
that, I think it's unlikely that I'm stripping unwind-related DWARF info,
but if there's something specific I should look for, I'm happy to do that.
If you don't have any other ideas, I guess I'll try going back to 3.6.0 and
patching out the atexit functionality like you suggested previously.
On Wed, Feb 11, 2015 at 4:51 PM, Jason Evans <jasone at canonware.com> wrote:
> On Feb 11, 2015, at 4:45 PM, Brock Pytlik <brock.pytlik at gmail.com> wrote:
> > I think these are the relevant lines from config.log:
> > configure:8897: checking configured backtracing method
> > configure:8899: result: libgcc
> > I see the same lines for both the x86 and arm builds.
> Unless you are stripping your binaries of all the unwind-related DWARF
> info, I don't have any likely explanations for the backtracing failures.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the jemalloc-discuss