PATCH: PR 42748

[Mark Mitchell]

Jason Merrill wrote:

>> This patch makes the mangler set the source location to the location
>> of the declaration while doing the mangling.
> 
> I've also been thinking about moving mangling messages to the decl locus
> instead, but by changing the warning call (as below) rather than by
> messing with input_location.

I don't think that's sufficiently general, at least without a bunch of
other changes.  In the case of the PR, the message comes from code in
the ARM back-end that has only a type, not a decl.  We'd have to find
some way to push the decl down.

>> I also made it push a
>> template instantiation level in the case that the manglee is a
>> template, so that we get the usual template instantiation backtrace in
>> that case.
> 
> Good idea.

So, do you think my patch is a good thing, or that we should pursue your
approach instead?

Thanks,

-- 
Mark Mitchell
CodeSourcery
mark@codesourcery.com
(650) 331-3385 x713