Subject: Re: Compiling NetBSD with another compiler.
To: None <email@example.com>
From: Trevor Talbot <firstname.lastname@example.org>
Date: 10/15/2007 12:16:20
On 10/15/07, Aleksey Cheusov <email@example.com> wrote:
> Even if lcc/tcc etc. are just a toy and doesn't provide results
> comparable with gcc, NetBSD user should (in my opinition) have ability
> to play with them and probably improve them in the future
> to support cleverer support
> for function renaming. But now all these compilers
> just don't work.
I personally don't see this as a problem. A toolchain's job has
always been to support development for a platform. Simply being able
to compile code for a particular CPU is not expected to be good
enough. A platform has, at least, various implementation-defined
conventions, be it symbol naming (underscore or not?), calling
conventions (use registers? which ones?), structure padding and
alignment, etc. Function renaming is just another convention, and
it's perfectly reasonable to expect toolchains that want to support
this platform to implement the necessary semantics.
Given NetBSD's position as being a portable platform in general, I'd
be unhappy if it were doing things differently just for the sake of
being different. Function renaming isn't gratuitous though, it fills
a specific need, so I don't see anything wrong with it.