r274246 - [codeview] Emit qualified display names if -gline-tables-only is on
David Blaikie via cfe-commits
cfe-commits at lists.llvm.org
Mon Jul 11 15:30:28 PDT 2016
On Mon, Jul 11, 2016 at 1:05 PM, Reid Kleckner <rnk at google.com> wrote:
> On Mon, Jul 11, 2016 at 10:35 AM, David Blaikie <dblaikie at gmail.com>
>> I believe so, but don't have specific numbers. Alexey made this choice
>> when it was originally implemented & I believe had the data back then.
> I don't think we made an explicit choice to use short names. gdb and
> addr2line just give you short names if you lack information about
> parameters, classes, and namespaces. I could be wrong though.
Not sure I follow - there was an explicit choice when creating the GMLT
mode to include only short names in our GMLT output, dropping namespaces,
linkage names, parameters, etc, I'm fairly sure (I seem to recall asking
Alexey before, pointing out that this produce surprising/confusing frames,
and him confirming that that was an intentional tradeoff for size).
> -- we can't use -fline-tables-only 'cause they make stacks look very bad,
>> Also I'm curious why your use case/tolerance for "badness" here is
>> different from what we've been using at Google for ASan, etc, for several
>> years now. Do you have different requirements/needs here? Then maybe we
>> need to figure out names for those needs & enshrine them in flags.
> I don't think our requirements are that surprising: users should be able
> to take a stack dump with a standard stack dumping tool (gdb, addr2line, or
> chrome's breakpad crash server) and get usable symbols (names that include
> their scope).
> Anyway, we all agree it requires measurement. I was mostly trying to gauge
> interest, to see if this is a problem that other users are encountering.
> I'd forgotten about Nico's RFC for -gmlt, so yes, it sounds like there is
> interest in revisiting this tradeoff.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the cfe-commits