Hello, OKUJI!
On Thu, 5 Aug 1999, OKUJI Yoshinori wrote:
> I have been thinking about the requirement for binutils. Most people
> have still been using ``old'' binutils, such as 2.8.1.0.x and
> 2.9.1. But we have recommened that the users should upgrade their
> binutils to a beta version. Although later binutils is definitely
> superior to older one because of the better support for 16bits code,
> it is never good to depend on any beta release.
IMHO:
It is Ok to require beta software to compile stable software.
It is Ok to require beta software to compile beta software.
It is not Ok to require beta software to use stable software.
> So I want to diminish the requirement as soon as possible. There are
> at least two possible ways:
>
> 1) Support any version of binutils. This means that we must assemble
> the 16bits code manually like grub-0.5.
> 2) Don't use gas for the 16bits code any more. Instead, use nasm or
> something appropriate.
"Something appropriate" is perhaps bin86, which is used by Linux and is
more widespread than nasm.
However, I would rather stick with binutils.
Compatability with more tools and compilers is good as far as it can
improve the program.
However, when such a compatability requires obfuscating the code or
rewriting it with a different syntax, this may become inacceptable.
Given the fact that asm.S uses both 16-bit and 32-bit code, using
different syntax for them opens the way for new bugs.
As long as binutils is free software, and no money hould be paid to
upgrade it, I see no reasons to save users (so called power-users, who
want to compile it) from making acquaitance with the new binutils.
Another questions is that the binutils maintainers should be contacted to
make a release. It is pretty disappointing to have new features for months
(if not years) and not to have them released.
Pavel Roskin
