linux-mips-fnet
[Top] [All Lists]

Re: Root disks

To: linux-mips@fnet.fr
Subject: Re: Root disks
From: Systemkennung Linux <linux@mailhost.uni-koblenz.de>
Date: Wed, 31 Jan 1996 21:45:54 +0100 (MET)
In-reply-to: <199601300138.SAA05659@rover.village.org> from "Warner Losh" at Jan 29, 96 06:38:49 pm
Ralf,

>       Thanks for putting together the Root disks!  I have a couple
> of questions...
> 
>       Is the source available yet for all of these files?

Of course, yes.  Most of the stuff are ordinary GNU programs.  For some
of them I needed some patches but nothing big.  Just include one headerfile
here and comment out line that conflict with libc declarations.  The
GNU libc is _very_ aggressive about this.  When I find the time I'll
clean up my sources and put the diffs on FTP.

                                                             It would
> be a good test of the system, once I get my machine booting, to try to
> rebuild them on Linux/MIPS.  If things aren't stable enough yet, I
> would understand.

All binaries except Binutils, GCC, Libc and the kernel itself have been
built under Linux/MIPS without problems.  For the mentioned programs there
are the following problems:

 - Binutils: I didn't try this yet.
 - GCC: Is too big to build with full optimization on a 8mb machine.
 - Libc: building libc does very much I/O and would probably take  a century
   on a diskless machine.  There is also some problem with signal handlers
   that breaks enquire.
 - Building the kernel fails because bash has some bug that makes it unusable
   for interactive use.  Nasty because the current shell ash is that
   user friendly.  (Hmmm ...  I should try my luck with the tcsh.  It's my
   favourite shell anyway ...)
 - Perl:
   Compile dies with a signal.  Might be that the compiler/binutils binaries
   use a ancient libc, might also be the above mentioned signal bug or even
   something completly different.

Since days I want to take the time to buy an IDE disk for my MIPS (SCSI
for the 53C94 is in the works btw.)  I've already tested this; with
lokal disk (-> buffer cache gets used) and swap the machine starts to
behave like I'd expect a R4400 machine to behave: *fast*.  Not a like
a topend machine but real fun to work with.

>       Also, would you be willing to take additional time in making
> these images so they aren't quite as large?  22M is about the limit
> that I can go over my link to the outside world w/o seriously
> negatively impacting other users.  If not, then I might be able to
> arrange for other channels to get these files, but I thought I'd ask.
> Are things to the point yet where this makes sense?

I admit that the size of the archive is far beyond the point where it is
still reasonable to distribute it as single file.  For the future I
want to use ordinary tar files for a single binary package which would
make upgrading easier.  As soon as I get the Debian scripts required
for package generation/installation running I want to start compiling
a real MIPS package.  My current problem is that Perl 5 doesn't build.

> Is it too early to talk about installation strategies yet :-)?

Definately not.  See above.

> Thanks again for the root disk...
> 
> That was the other thing... I had asked a while ago if I can put all
> these files on a disk created with Linux/Intel ext2fs and then hope to
> be able to mount that disk correctly with Linux/MIPS.  I don't recall
> seeing an answer coming back...

Yes, the filesystems are compatible.  If real live should proof them to
be incompatible it's a bug.

   Ralf

<Prev in Thread] Current Thread [Next in Thread>
  • Root disks, Warner Losh
    • Re: Root disks, Systemkennung Linux <=