linux-mips-fnet
[Top] [All Lists]

Re: Anybody compiled MILO on a non-linux box?

To: linux-mips@fnet.fr
Subject: Re: Anybody compiled MILO on a non-linux box?
From: Warner Losh <imp@village.org>
Date: Wed, 31 Jan 1996 09:45:47 -0700
In-reply-to: Your message of Wed, 31 Jan 1996 13:22:13 +0100
: Hum,... for now I haven't compiled Milo against glibc include files
: (which are the ones in Ralf's root package) I better check that !

I've long wondered why *ANY* of the library include files are used.
They are just fiction, since MILO doesn't use the library.  I think
most of them will be solved by not compiling -D__KERNEL__, a thought
that didn't occurr to me until this morning...

: I had never tried mipsel-linuxelf tools chain to compile Milo, but yes
: -oformat ecoff-littlemips should have done the trick as you say, yet
: another thing I should test...

I'll have to build the mipsel-linux tools here and give compiling
another spin.

: Warner> And the prebuilt MILO dies badly on my Deskstation.  TLB miss
: Warner> exception (Store).  The BADVADDR is 0x7ffffffb, which smells a
: Warner> lot like the new tagged code that Stoned has been working on
: Warner> (the EPC is right for that as well, if I read my nm's
: Warner> correctly).
: 
: Hum,... the distributed binary is compiled with the options not
: commented in the Makefile - if this help...

It would if ld didn't dump core on me :-)

: And more we don't have RAM size detection for the Deskstation it's fixed
: at 8Mo, you're welcome to write something similar to the
: magnum_type_bi() function (in magnum.c) wich sould detect the RAM size
: and add a pointer on this function in machinetbl[] element
: corresponding to the rPC44 (identifyc.). This should be something like:
: 
: struct machine_struct machinetbl[] = {
:  { "DESKTECH-ARCStation I", (char*)NULL, "DeskStation rPC44", 
:     dsk_rpc44_bi, (machine_func*)NULL,
:     taglist_deskstation_rpc44
:   },
: ....
: }

OK.  I'll patch it to at least be 32M, which is the default for my
machine :-).

: Warner> The video address for the RPC44 is incorrect.  It should have
: Warner> its own address, not share the TYNE one, and that address
: Warner> should be 0xa00a0000.  I'll see about hacking that into MILO
: Warner> and generating a patch, once I figure out how to build MILO
: Warner> :-(.  Also, I think the load address for MILO needs to be
: Warner> different given some of the comments in MILO's makefile.
: 
: Well just correct the vram_base tag value in taglist_deskstation_rpc44[]
: (identify.c).  

I have a patch, I think, that I can send you for that.  

: Well, all this identification stuff ins't very usefull if Milo keeps
: crashing but...

Yes.  I'll look more into this tonight...

Warner

<Prev in Thread] Current Thread [Next in Thread>