At 17:22 -0700 1998-07-23, Dan Jacobowitz wrote:
>[Before you read my response, one comment]
>I'm not sure you're taking a great approach here. There's a very well
>tested set of scripts in boot-floppies for creating a debian base
>tarball. I intended to use those when the last missing binaries are
>dealt with.
I've done some preliminary work in this area, I'm far enough along to
probably make the base tarball. We can start with the base tarball and the
Mac OS "Installer" the Linux/mac68k folks wrote, it understands ext2fs, and
has GNU tar and a bunch of other programs built into it (it's also slower
than hell in disk I/O). There is also a Mac OS port of the e2fsprogs, which
will be needed to make the filesystems. Hmm... perhaps I'll write up an
INSTALL doc (once I have a tarball to test with).
(as an aside, an evil thought just struck me, sticking the a filesystem
image of the whole tarball in as the ramdisk.image.gz for the kernel
makefile and booting it, I figure it'd fit in my 64MB of RAM with enough
room left-over to run things, this is assuming OF on a PowerMac can handle
booting a coff image off a HFS formatted HD partition <fx: maniacal
laughter...>).
>I'm not sure how often ftp.debian.org/pub/debian/Incoming updates.
Last I checked, ftp.debian.org doesn't mirror incoming. llug.sep.bnl.gov is
a good mirror of incoming (with bo gone, it probably has space for the full
archive again too (source was removed a while back, and slink wasn't
mirrored)).
>with getty, more, and hwclock (still broken), it's in util-linux.
hwclock just barely supports non-i386 architectures, i.e. first it tries
some ioctl, and if that fails, it tries direct hardware access, this is so
it works for m68k and possibly alpha besides i386. The standard Debian init
scripts are designed to cope with either a 'clock' binary or a 'hwclock'
binary, since pmac-utils has a 'clock' binary, we should be able to use it.
>That's the meaning of Essential: more than base/ AFAIK. netstd I just
>uploaded; netbase is working in slink I believe.
This is why I find base annoying, the contents of the 'base system' don't
even match the contents of 'base'.
I proposed ditching the 'base' section in favor of using the priorities
field to determine what's part of the base system on IRC to a few people,
but no one else liked my idea.
Thinking about it, I think I will propose that base not be a "real"
section, instead, it should be symlinks to the actual packages that make up
the base system (perhaps with a new priority value; "base" to go with it).
>> Missing Binaries snarfed from RedHat: (all libc1.99, so they don't work)
>I deleted ones I have already accounted for.
>> /sbin/kbdrate
> It's there...it doesn't seem to DO anything though. This is also a
>util-linux program, and I'm still working on it.
It might be dependent on PC-like keyboard hardware (some non-PowerMac ppc
systems might have such hardware, for all I know), and only compiles on ppc
through luck :).
>> /usr/sbin/cytune
A util for Cyclades multi-port serial cards, could work on a PPC system
with ISA slots, assuming appropriate card.
>> Missing Binaries which were removed: (didn't have 'em in RedHat)
>> /bin/ae
Odd, I was sure I'd built that. It needs patching to compile under glibc
2.1, but there's a patch in the bug tracking system from the sparc folks.
<<
diff -Nur old/ae-962/header.h ae-962/header.h
--- old/ae-962/header.h Tue May 19 18:21:00 1998
+++ ae-962/header.h Tue May 19 18:11:54 1998
@@ -250,7 +250,9 @@
extern void msg _((t_msg, ...));
extern char *getmsg _((t_msg));
extern char *strlwr _((char *));
+#if !((__GLIBC__ > 2) || ((__GLIBC__ == 2) && (__GLIBC_MINOR__ >= 1)))
extern char *strdup _((const char *));
+#endif
extern char *strrep _((char *, int, int, int));
extern char *pathname _((char *, char *));
extern FILE *openrc _((char *));
diff -Nur old/ae-962/main.c ae-962/main.c
--- old/ae-962/main.c Tue May 19 18:21:00 1998
+++ ae-962/main.c Tue May 19 18:12:05 1998
@@ -487,6 +487,7 @@
return (str);
}
+#if !((__GLIBC__ > 2) || ((__GLIBC__ == 2) && (__GLIBC_MINOR__ >= 1)))
/*
* Make a duplicate of a string. Return a pointer to an allocated
* copy of the string, or NULL if malloc() failed.
@@ -500,6 +501,7 @@
(void) strcpy(new, str);
return (new);
}
+#endif
/*
* Replace old with new characters. If method
>>
>> /bin/fdflush
fdflush is for *really* cheap PC floppy drives, most powerpc hardware is of
high enough quality not to need it.
>> /sbin/activate
part of lilo, ax it
>> /sbin/cardmgr
>> /sbin/cardctl
>> /sbin/ifport
>> /sbin/scsi_info
>> /sbin/probe
>> /sbin/ftl_format
>> /sbin/ftl_check
pcmcia^WPC Card stuff from pcmcia-cs, dunno if pcmcia is supported on PPC
notebooks yet.
>> /usr/bin/elvis-tiny
>> /usr/bin/resizecons
resizecons is i386-specific
>> /usr/bin/syslinux
May be usable for PReP/CHRP machines (syslinux creates bootable FAT
formatted floppies). m68k (except Mac), i386, and alpha use it for boot
floppies, AFAIK.
>> /usr/sbin/tunelp
Parallel-port specific, IIRC. (non-PowerMacs might find it useful).
>> /usr/sbin/chat
>> /usr/sbin/pppd
>> /usr/sbin/pppstats
'ppp' part of the base system, but isn't in 'base' (see my rant above ;)
>> /lib/ld.so
libc5 dynamic linker
>> /usr/lib/lddstub
powerpc uses glibc's ldd, so that can go.
>> /usr/lib/perl5/i386-linux/5.004/auto/IO/IO.so
>> /usr/lib/perl5/i386-linux/5.004/auto/Socket/Socket.so
One of the dselect methods probably needs those.
>> /usr/lib/libnewt.so.0.21
>> /usr/lib/whiptcl.so
newt0.20 and whiptail (used by dinstall on the boot-floppies)
--
Joel "Espy" Klecker Debian GNU/Linux Developer <mailto:[EMAIL PROTECTED]>
<http://www.espy.org/> <ftp://ftp.espy.org/pub/>
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
