Bruce Dubbs wrote:
> If that were the policy at one time, then it should be abandoned.  We now 
> require sed 3.02 as a part of the host requirements.  According to the change 
> log, the in-place mechanism has been in sed since 2001-12-17.  Looking at the 
> sed releases, that is about the time that sed-4.0 was released.

Ah, well, then there ya go.

> It would be much easier to change the host requirements to sed-4.0.6 
> (18-Mar-2003) or later rather than bend over backward to accommodate a host 
> that 
> has a package that predates the 2.6 kernel.  I chose 4.0.6 as a benchmark 
> because that is the earliest sed-4.0 in the gnu repository.
> 
>> (A complete aside is that we could also always do a very early build of 
>> sed before anything else, and bash and make for that matter, like DIY 
>> does so that the entire build process is more robust, is buildable on 
>> older systems and we can use more 'modern' commands throughout the build.)
> 
> We already require bash-2.05a and make-3.79.1 or later.  Are there issues 
> with 
> these?

There's no real issue, more just an opinion. I personally don't care so 
much that we have 'requirements' that aren't really requirements. In 
other words, it is possible to build LFS (with a few slight 
modifications) on hosts with older software than we specify, so I don't 
really see a need to require more modern versions especially if it's 
just so we can use more convenient syntax.

A 2.6 kernel so that glibc can build itself and its nptl libs in a sane 
way is a technical requirement. Sed-4.0.6 so that we can use '-i' is not.

As I said, if you want to use -i, great. The build ends up being more 
robust and you have less specific requirements for the host if you build 
sed twice in chapter 5 (once before you do binutils) and you get the 
bonus of being able to use -i.

--
JH
-- 
http://linuxfromscratch.org/mailman/listinfo/lfs-book
FAQ: http://www.linuxfromscratch.org/lfs/faq.html
Unsubscribe: See the above information page

Reply via email to