Re: gcc4 problem on netbsd-4

Ok, I just tried this kernel again:
http://www.ceres.dti.ne.jp/~tsutsui/netbsd/netbsd-ATARITT-4.0_STABLE-20081231.gz

And this time I used the sysinst from:
ftp://ftp.netbsd.org/pub/NetBSD-daily/netbsd-4/200812280002Z/atari/installation/miniroot/

Sysinst runs, but I get a read-only filesystem.  I see:
mount_ffs: realpath /kern/rootdev: Bad file descriptor
cp: /tmp/.root_writable: Read-only file system

Then I can run mount and see that the root filesystem is read-only.  This 
appears to be the bug that David Brownlee fixed in HEAD:
http://archive.netbsd.se/?ml=port-atari&a=2008-11&m=8974408

I guess the fix didn't make it into netbsd-4.  I think David's changes will 
need to be submitted to netbsd-4 in order for me to proceed further in 
testing this kernel.  This would be helpful also because then the 4.1 
release may be functional on Atari when available.

David Ross
[email protected]


----- Original Message ----- 
From: "David Ross" <[email protected]>
To: "Izumi Tsutsui" <[email protected]>
Cc: <[email protected]>; <[email protected]>; <[email protected]>
Sent: Friday, January 02, 2009 10:29 AM
Subject: Re: gcc4 problem on netbsd-4


>> BTW, did you try the latest netbsd-ATARITT-4.0_STABLE-20081231.gz one?
>> If so it's a good news that all devices are probed properly
>> on a 4.0_STABLE kernel and TT030 :-)
> Yeah, this is the one I tried.  But wait, actually, should I be using the 
> sysinst from NetBSD-daily/netbsd-4 or NetBSD-daily/netbsd-4-0?  (I'm not 
> sure what the difference is.)
>
>> Of course the HEAD kernels have more new features so
>> they are ever growing.
> But the wierd thing was that I believe it's the 4.0_STABLE one that's the 
> larger kernel.  But I guess if your HEAD kernels were gcc3 (?) and the 
> 4.0_STABLE one is gcc4, then that would explain it.
>
> Next things on my agenda to try:
> - The 4.0_STABLE kernel with the correct sysinst image
> - A HEAD install with AHDI-compatible partitioning (this might avoid the 
> bootloader problem)
> - The new ATARITT kernel on the "wdc and DMA interrupt issue" thread
>
> David Ross
> [email protected]
>
>
> ----- Original Message ----- 
> From: "Izumi Tsutsui" <[email protected]>
> To: <[email protected]>
> Cc: <[email protected]>; <[email protected]>; 
> <[email protected]>
> Sent: Thursday, January 01, 2009 3:44 PM
> Subject: Re: gcc4 problem on netbsd-4
>
>
>> David Ross wrote:
>>
>>> On the TT030, it boots but after loading sysinst I get a looping 
>>> printout
>>> of:
>>> erase ^H, werase ^W, kill ^V, intr ^C
>>> mount: /: Bad system call
>>> erase ^H, werase ^W, kill ^V, intr ^C
>>> mount: /: Bad system call
>>> ..
>>>
>>> I think probably the issue is that I need to switch from using the HEAD
>>> sysinst to the netbsd-4 sysinst.  I'll try this later.
>>
>> Yes, binaries for a newer release might use newer system calls
>> which older kernels don't know.
>>
>> BTW, did you try the latest netbsd-ATARITT-4.0_STABLE-20081231.gz one?
>> If so it's a good news that all devices are probed properly
>> on a 4.0_STABLE kernel and TT030 :-)
>>
>>> The filesize on this kernel is different from the others -- is this one 
>>> gcc4
>>> whereas the others were gcc3?  I thought HEAD was already gcc4 (?)
>>
>> Even 4.0 uses gcc4 by default, and I just built gcc3 kernels
>> (with -gcc3 suffix) to see if the long delay problem
>> (it was first reported in PR port-atari/37464) was caused
>> by gcc4 or other changes.
>>
>> Maybe gcc4 is more aggressive on various optimization
>> (more inline functions, align instructions/data at cacheline etc.)
>> and produce larger binaries (unless -Os is specified).
>>
>> Of course the HEAD kernels have more new features so
>> they are ever growing.
>> ---
>> Izumi Tsutsui
>>
>