Bug#31450: boot-floppies 2.1.4 fails to build for m68k
First off, my apologies for not noticing the messup Netscape caused
for me. I've reassigned the bug now, and informed Ian about this nice
feature.
Adam Di Carlo wrote:
>
> Michael, first off, care to help with the documentation of the
> Installation System? Cf
> http://www.debian.org/~aph/boot-floppies/m68k/install.html . I need
> the m68k stuff to be critiqued, corrected, and filled out. You can
> send me text in ASCII or SGML (linuxdoc), I don't mind.
For starters, check http://www.linux-m68k.org/debian-{atari,mac}.html.
There's Franky's Amiga install guide out somewhere too, I forgot the
URL. They walk users through the details of the actual install but don't
cover the system requirements and other things in section 2 of your
document.
The machine specific install guides available would fit in section 6
I guess. I only had a brief look at them but will have a closer look
(wish I could get it as ASCII text :-)
> > - kernel-image-2.0.35 doesn't exist for m68k, substituting 2.0.33.
> > Eventually, 2.0.36 will be released for m68k, just FYI.
>
> You can set whatever kernel you like in the top-level makefile. How
> is this a bug?
Yea, that's right, it's a feature. Hardcoded kernel versions later on
are worse. And the unavailability of a Mac kernel image package will
force me to fake one in the end.
I guess I was complaining that these m68k features are still not known
or
considered in the boot-floppies team. I admit that I started out without
reading the README :-)
> > - rescue.sh expects to find 'root{atari,amiga}.bin' ramdisk images
> > but fails to pass 'atari' or 'amiga' as argument 1.
>
> Oh, that's bad.
It gets even better; both ramdisks and base system have an incomplete
set
of device files in /dev. Has all sorts of funny side effects. Patch (for
part
of the problems, untested on base yet) and description in a separate
followup.
> > - scripts/basedisks/master2files/master contains 27 packages that
> > don't exist for m68k, plus kernel-headers-2.0.35 (replaced with
> > 2.0.33, see above) and lsof (also required for 2.0.33).
>
> Woah, we should find a way to not hard-code the kernel-headers entry.
>
> > A number of these packages are Intel-specific (i.e. X servers) and
> > will never be available for m68k. A number of other packages are
> > broken and can't be built for m68k, compare
> > http://debian-m68k.nocrew.org. The packages:
>
> > -acidwarp -bin86 -checker -clisp -fdflush -ftape-module-2.0.34
> > -ftape-util -gcl -guile1.3 -libguile4 -libguile4-dev -lincity-svga
> > -lockvc -lsof-2.0.35 -luxman -netdiag -omniorb -razzle -svgalibg1
> > -vgacardgames -vgagamespack -xemacs20-bin -xemacs20-nomule
> > -xserver-mach64 -xserver-s3 -xserver-svga -xserver-vga16 -zgv
>
> > For the above, see attached diff (as a temporary fix to permit
> > testing of the 2.1 base system on m68k).
>
> Thanks. I'm not sure how we're coping with this problem of
> architecture-dependant lists. I would suggest we use either cpp or m4
> to auto-expand the list at build time based on the arch.
Either that or use arch-dep master.<whatever> and make master2files
smarter (so it reads the name of the input as argument). I don't care, I
patched master2files to ignore all of this and produce the output I need
to get to the good stuff.
> > - a number of packages had incorrect dependencies (not in same task,
> > not in same profile). I can supply a detailed log if wanted. These
> > dependencies result in failure of the build even using
> > --nodependencies_errors_are_fatal for master2files (no patch, I
> > assume that's eventually fixed before release).
>
> Interesting. I would think this would have been fixed prior to any
> release...!
See files test.master and depend.master in ~schmitz/boot-floppies,
that's
the complete list after 'immunizing' master2files.
> > - my main gripe with boot-floppies: even with check_mirror.sh
> > there's still some mysterious dependencies left:
>
> > * ../bootstra.ttp (what's with the Amiga bootstrap, anyway???)
>
> No clue!
It's missing. Question goes to Franky: I'll extract the booter from the
2.0 set if necessary, ok?
> > * recode * libc6-pic * recent makedev
>
> What's the issue here? I believe these are indeed required to build.
They are - is it mentioned somewhere, did I miss something? Didn't show
in
James' source dependencies, at least.
makedev (slink/binary-m68k as of yesterday) isn't recent enough, it
doesn't know generic-m68k (or generic, for that matter: 'generic-generic
not known'. Duh.).
Something strange happens at the end of the whole install (after adding
some
of the devices required manually to speed up the thing): I can't log in;
the login never prompts for the password but claims incorrect login. Any
idea on this?
I'll see that I get a new set of base and rescue disks out tonight;
don't
touch the current set, it's totally broken. New build in progress.
Michael
Reply to: