2.4.1-Balu1 pre-chroot (LONG)

Thomas 'Balu' Walter tw at itreff.de
Thu Oct 12 11:47:21 PDT 2000


+-Gerard Beekmans-(gerard at linuxfromscratch.org)-[12.10.00 13:46]:
> On Thu, 12 Oct 2000, you wrote:
> > Okay - here is my up-to-chroot-description of 2.4.1...
> 
> Uh oh ;)

> Again Balu your reports are invaluable to me. Glad you take the time almost 
> every release to go through it.

8) I'll give my very best :)
> > Whish from me (to be discussed):
> > Get the prompts back (I don't like the style with &&... - even if I use
> > it usually, it looks wierd. I agree that cutnpaste is easier then and I
> > see that you have less to edit while updating packages ;)
> 
> Those are my main reasons for removing the prompts yes. I agree it looks 
> prettier with the prompts in place, but I believe the removal of them is 
> worth it. I'm reluctant to add them again, unless there are of course a lot 
> of people who want them back. If you want, propose a way to make copy&paste 
> as user friendly as it is now.

Okay - forget the prompts, but what about deleting the &&s at the end -
I am not sure and can not try it at the moment, but I remember having
cutnpasted multiple lines including returns (this works like typing a
command if another is not ready yet).

But I am not sure if this is a nicer and recommended way...

> > Installing Bash:
> > 	failed first, after installing libncurses5-dev and
> > 	libncurses5-dbg it worked
> > -----
> > install-info: unknown option `--dir-file=/mnt/lfs/usr/info/dir'
> > usage: install-info [--version] [--help] [--debug] [--maxwidth=nnn]
> >              [--section regexp title] [--infodir=xxx] [--align=nnn]
> > 	     [--calign=nnn] [--quiet] [--menuentry=xxx] [--info-dir=xxx]
> > 	     [--keep-old] [--description=xxx] [--test] [--remove] [--]
> >              filename
> > make[1]: *** [install] Error 1
> > make: [install] Error 2 (ignored)
> 
> That's because Debian's texinfo package is hopelessly outdated.

:( What goes wrong then - the info-pages are not installed?
> 
> > Installing Bzip2:
> > 	Reading the man-page I think "make -f -" would be the correct
> > 	way to get the Makefile from stdin.
> 
> It's my opinion that -f /dev/stdin is clearer than -f -
> 
> Both are ok (since that's just what make does anyways - use /dev/stdin if you 
> provide it with a dash)

On that Sparc IPC I was playing with (and which is somehow broken at the
moment - no time to look at it) I had the problem that /dev/stdin did
not work, don't ask me why, but changing it to - helped. And since the
man-page says "-" for /dev/stdin I thought we should replace it...
> 
> > Installing Linux Kernel
> > 	I took 2.2.17 instead of 2.2.16, which is fixed in 2.4.2 now.
> > -----
> > /mnt/lfs/usr/src/linux/include/linux/coda_opstats.h is empty
> > /mnt/lfs/usr/src/linux/include/linux/dasd.h is empty
> 
> Not to worry about i think...but I'll keep an eye on it next time I build an 
> LFS system (tomorrow)

I did not worry about that really, but I was wondering why the
kernel-developers create empty include-files ;)
> 
> > 	I had problems cutnpasting from w3m here - spaces were wierd
> > 	characters and were not displayed after cutnpaste - I looked
> > 	into the html-source and found  es - why are they used
> > 	here?
> 
> If I don't add the   (Non Breaking SPace) Netscape or most other 
> browsers refuse to put extra spaces. It might insert one space but ignores 
> extra whitespace. Because the commands are broken up over multiple lines, 
> indenting them so you can see that they belong to one and the same command is 
> better. That can be done with  . No browser should give you weird 
> characters, this stems back from the HTML-1.0 era and should not pose any 
> problems....got a proposal to fix it perhaps? 

I agree, but then I thought that w3m (and lynx?) does that to avoid the
sentences being wordwrapped at the end of the line - ASCII-255 is just
another space-character, which the shell does not like though...
> 
> > 	What about creating configparms with:
> > 		cat <<EOF >configparms
> > 		# Begin configparms
> >
> > 		slibdir=/lib
> > 		sysconfdir=/etc
> >
> > 		# End configparms
> > 		EOF
> > 	The echo-line came too late ;-)
> > 	make is too fast - I can not see if I get "character 45" errors...
> > 	Is it bad to apply the patch anyway? I did not - and I think I
> > 	did not have the errors (00-Errors didn't show them)
> 
> No, it won't hurt to apply the patch. I'll add a line to find out what bash 
> version is being used so you can anticipate the need of the patch before you 
> start building LFS. That sounds logical doens' it? Then you might wonder why 
> I never added it before. I wonder it too. I know it has crossed my mind on 
> several ocassions...

=) - I was not sure why that patch had to be done anymore...
> 
> > Creating passwd and group files
> > 	Create the group-file like configparms above?
> 
> Interesting idea, I can add the echo line underneath it like I did with 
> configparms now.

the echo-lines are not needed anymore then (create the passwd with echo)
and the group with <<EOF... -
> 
> > Debugging symbols and compiler optimizations
> > 	What do I take for Thunderbirds? -mcpu=?
> > 	The gcc-info pages say that `-march=CPU TYPE' implies `-mcpu=CPU
> > 	TYPE'. So we don't need -mcpu=?

Do we need -mcpu?

> > Creating $LFS/root/.bash_profile
> > 	Again - create that "cat <<EOF >file"-style?
> > 	Ah - here we have "yyy"...?
> 
> Ah there the yyy's went to, I knew they were supposed to be somewhere.

Yes, but they are not needed, because the same value is used both times
;), and anyway - -march implies -mcpu, so leave that out.
> 
> Sure, the echo (or cat version) can be added below the contents of the file.

No need to echo then - the file is created with that statement... (and
no extra writing of the file-content).

     Balu

-- 
Unsubscribe: send email to lfs-discuss-request at linuxfromscratch.org
and put unsubscribe in the subject header of the message




More information about the lfs-dev mailing list