jakelly at shtc.net
Tue Sep 27 10:35:52 PDT 2005
On Mon, 26 Sep 2005 22:07:32 -0400, John Kelly <jakelly at shtc.net>
>If you can mirror his technique when extracting the build commands
>from lfs, then I can readily plug them into my scripts.
Greg puts the commands in a directory named "scriptlets" and creates a
subdirectory for each chapter, namely "temptools" and "chroot," but
the subdirectory names may be chosen arbitrarily. Whatever the names
are, I can map them into corresponding stages like so:
stage1 chapter 4. Final Preparations
stage1.src chapter 5. Constructing a Temporary System
stage2.src chapter 6. Installing Basic System Software
All stages are handled by the same wrapper, so the difference between
a stage? and its corresponding .src is arbitrary. However, the intent
is to use .src stages for the principal tarball build scriptlets, and
each corresponding stage? for preparatory environmental scriptlets.
A diy-linux anomaly is that Greg dumps his chapter 4 environmental
commands directly into the top level "scriptlets" directory, but that
works anyway, because I map them into stage0 (though I could just as
well use stage1 for that purpose).
In diyl-0.002.tgz, you may notice that scriptlet and tarball names are
"hard-coded" in the various stages, and wonder how that can work with
Greg does not assign any sequence numbers to his diy-linux scriptlets,
so you have to map them, by hand, into the proper sequence, and that's
what those scripts do.
Though problematic, that requirement also proved beneficial, because
the stage scripts thus provide an abstraction layer, letting the user
customize his build as he pleases.
However, because the stage?.src scripts all have a common, regular
structure, instead of hand crafting them, I can generate them by
machine, provided that scriptlet sequence numbers are available for
that purpose (as with jhalfs).
That's how it can work.
More information about the alfs-discuss