[llvm-dev] lld: ELF/COFF main() interface

[Rafael Espíndola via llvm-dev]

> After re-reading the thread, I also believe that Chandler could have worded
> some of his replies better (for example, his comment on "Rafael introduced
> error() function later and we now depends on that function does not return."
> was "I think the last was a mistake"). If Rafael was hurt / offended by one
> of his technical decisions being outright called a mistake, the right way is
> to raise this in a calm way - I'm sure Chandler would be happy to apologize
> for his wording if this is the root cause for this argument.

I consider LLD to be the most important software I have worked in. I
was very much hurt by the suggestion that we were going in a opposite
direction from what was discussed. The discussion and decision on what
the current design is is what motivated me to join the project. The
original patch was already clearly not a library, so it was there for
everyone to see.

I was also very much hurt by the suggestion that it should not be part
of LLVM if it doesn't match Chandlers' preferences, regardless of  how
much code he has in it. If that is really the case this project
structure is now broken and I am going too.

So I stand that if anyone wants a library, it is up to that person to
figure out the API, design and send a patch.  It is also important to
provide real reason. "a fork in an entire build is too much" would
need a lot of data to justify.

A big part of the discussion about LLD occurred in a meeting with
Chandler present. The fact that that discussion faded out is the main
reason I think It is a bad idea to discuss in place without a log.

Cheers,
Rafael