On 03/11/2017 07:41 PM, Jeremy Huntwork wrote:
Hello,
I was just going through the latest book again as a reference/review.
Nice job on keeping it up to date and valid!
There were just a couple of small things I noticed which could possibly
be improved.
In section 5.5, Gcc pass 1, the following switch is not explained:
--with-glibc-version=2.11
I'll add a description at next commit. Probably coreutils-8.27-i18n.
Oh, FYI, unless there are objections, think I'm going to wait a bit and
use mbfile exclusively for this iteration of the coreutils-i18n patch.
The separate functions (from Suse) have been pulled forward a few times,
and ultimately, despite the much cleaner looking patch set, it's going
to be more difficult to maintain in the long run vs using the mbfile
addition (courtesy of RedHat - which we are already using for sort and
expand).
In section 5.8, I'm curious why --host= was chosen for the libstdc++
build instead of manually specifying the compiler and utils to use as is
done for binutils pass2 and gcc pass2.
I imagine at that stage it would
be better to let the build script assume it is compiling fully natively
with a manually specified compiler and utils instead of possibly trying
to do something special with cross-compiling. I'm sure everything
currently works, it just seems inconsistent and possibly open to
unwanted effects. If there is a strong reason for using the --host
switch, a note about why would be helpful.
Actually, it's probably just a leftover -- it still works as is. For
gcc-5.x, IIRC, both --host and --target were required. Will try to
address next week if one of the other guys doesn't beat me to it.
It's also not explained why libstdc++ is done separately now before gcc
pass 2. From going through it, it seems that with version 6.3.0 it's a
hard requirement, but a brief note about why it is built separately
beforehand instead of just later as part of gcc pass 2 could be helpful.
Suggestions for text are welcome. If not, we'll get something in there
to address it.
There was another thing I had noticed in the stable book, about the
names of the gcc headers to adjust for the startfile prefixes, but it
appears you've fixed it already for the development one, so good job!
Yeah, that one was totally my fault. I dropped it at some point in the
patch I was carrying forward for several months before the lib path
changes went in, It was only fixed last night.
Thanks for the observations. We will get them addressed shortly.
--DJ
--
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
