Re: native-comp-async-report-warnings-errors default value

[Andrea Corallo]

Dmitry Gutov <dgutov@yandex.ru> writes:

> Hi Andrea,
>
> On 20.12.2021 11:29, Andrea Corallo wrote:
>> Stefan Kangas <stefankangas@gmail.com> writes:
>> 
>>>> Eli Zaretskii <eliz@gnu.org> writes:
>>>>
>>>>> When I said "by popular demand", I meant that.  Just your own voice is
>>>>> not enough, IMO.
>>>
>>> I have now reviewed this thread, and changing the default from nil to
>>> silent has the support from Robert Pluim, Stefan Kangas, Lars
>>> Ingebrigtsen, Andrea Corallo and Óscar Fuentes (5).
>>>
>>> For keeping the current default I count Eli Zaretskii (1).
>>>
>>> Neutral in this thread are T.V. Raman and Stefan Monnier (2).
>> FWIW please count me on the neutral side, I've no formed opinion on
>> this
>> ATM even if the more time is passing the more I tend to think that the
>> current default might be safer and/or beneficial on the long run.
>
> This might be a distracting aside, but could you comment on the idea
> that I think Stefan mentioned previously: that the native compiler
> could use not the original source files, but the byte-compiled *.elc
> files as the "source". By the time those are compiled, all the
> dependencies are resolved one way or another, right? So the user won't
> need to suffer the same kind of warnings again.
>
> Do they contain the necessary information to do native-compile, and
> its various optimizations? If yes, are there any other difficulties
> associated with such an approach?

Hi Dmitry,

this is not a trivial modification so I might easily miss some aspects
of it but...

Certanly bytecode ATM has no way to store declaration specifiers and
these are used for native compilation.  I guess we should at least
extend this first.

BR

  Andrea