| Newsgroups |
gmane.linux.lfs.beyond.devel |
| Message-ID |
<174141457295.8.14744255912673435012.628324435@4forl1st5.slmail.me> |
On Saturday, March 8th, 2025 at 04:47, Xi Ruoyao <[email protected]> wrote:
>
> On Sat, 2025-03-08 at 12:31 +0800, Xi Ruoyao wrote:
>
> > On Sat, 2025-03-08 at 03:55 +0000, lfs wrote:
> >
> > /* snip */
> >
> > > Should that command be rendering in a way that would see JHALFS try to
> > > execute it in the first place?
> > >
> > > The source (postlfs/security/make-ca.xml) has
> > >
> > > <screen role="root"><userinput>/usr/sbin/make-ca -g</userinput></screen>
> > >
> > > which is why it's getting dumped, but maybe it should be
> > >
> > > <screen role="nodump"><userinput>/usr/sbin/make-ca -g</userinput></screen>
> >
> > No. We deliberately don't use nodump here.
> >
> > If we change it to nodump, packages "depending on" make-ca will fail to
> > build because in fact they depend on the certificate store to be set up,
> > not just make-ca.
>
>
> I.e. if we change it to nodump, someone will get a FTBFS if (s)he
> selects Mesa in jhalfs. And then (s)he'll complain "LFS/BLFS Books
> aren't written so as satisfy JHALFS."
>
> The problem is not "LFS/BLFS Books aren't written so as satisfy JHALFS."
> It may be true in the sense LFS/BLFS are supposed to be read by a human,
> but then JHALFS is supposed to be adjusted to satisfy LFS/BLFS. I.e. if
> jhalfs fails for a LFS/BLFS package we should either fix LFS/BLFS (if it
> indicates a real bug in LFS/BLFS), or fix jhalfs.
>
> Here the problem is some BLFS packages are simply not supposed to be
> build offline (no matter with or without JHALFS) and we don't want to
> spend our time on implementing such a support. And making this
> role="root" instead of role="nodump" is EXACTLY for satisfying jhalfs,
> not the opposite.
>
> If you just select some random packages without reading the book to see
> if building them (and their dependencies) needs an Internet connection,
> you should NOT assume the build can be totally off-line.
Hardly random, given I know that I can build that payload off-line,
baring the "run time" dependency that I know I can leave until later.
But in general, yes, I understand what you are saying there.
So I think the thing that I've missed, about BLFS, was in Bruce's
statement, echoing your last line, that
"the book is written assuming network connectivity."
something which, in all the time I've been using (B)LFS, I've never
really absorbed having always preferred to have pulled down sources,
ahead of building packages, and that includes since I've started to
use JHALFS, whereby I'm typically pulling down ALL of the LFS and
BLFS packages and patches I know I'll need, on the old "host system".
As for Bruce's note,
"In BLFS-12.3 there is no longer an unzip package."
I was aware of that - as I do read the git logs - so just to repeat
that the list of scripts I'd presented, for context, was taken from
my old notes, so from a pre-12.3 install.
Thanks for taking the time to respond, especially as this has turned
out, once again, to be "my issue" in "my distro" and so something I
will continue to fix up in "my rules".
--
http://lists.linuxfromscratch.org/sympa/info/blfs-dev
Unsubscribe: See the above information page