Re: [PATCH] Add --devel-skip-producer

[Tom de Vries <tdevries@suse.de>]

On 3/17/21 12:14 PM, Mark Wielaard wrote:
> Hi Tom,
> 
> On Wed, 2021-03-17 at 11:46 +0100, Tom de Vries wrote:
>> On 3/16/21 10:27 PM, Mark Wielaard wrote:
>>> It is useful as --devel option, but I would be against it as non-
>>> devel
>>> option.
>>
>> Could you explain in more detail why you would be against this as a
>> non-devel option?  F.i., I'm curious, is it an abstract objection, or
>> do
>> you foresee concrete problems with the approach?
> 
> Of course I would never call my own objections abstract :)
> But it partly is. I think it is questionable if we have to rely on the
> producer string to do the right thing.

I'd say dropping a CU is not the right thing, it's a workaround, and
relying on the producer string for a workaround sounds acceptable to me.

> But specifically just dropping a
> whole CU DIE tree seems the wrong thing to do (what about the ranges,
> loclists, stmt_lists left behind for example).

Yeah, the feature may not be complete yet. F.i. if there's a dwarf
expression in a loclist referring to a DIE in the dropped CU, then
that'll blow up probably.

But for dwarf generated by assemblers, which is not that complicated,
this seems good enough.

Thanks,
- Tom