[rescue] Sunshack and Sun stuff to giveaway

[Shannon Hendrix]

On Oct 29, 2007, at 12:58 PM, der Mouse wrote:

>>> Well, I'm still interested in SS20s; they're the best (FSVO "best")
>>> Sun that's actually usable.  I've got some Ultras, but I ran into a
>>> bug that looked to me like a gcc sparc64 codegen bug
>> That seems a silly reason to not use Ultra hardware.
>
> Sorry, but-- I don't feel any particular compulsion to limit myself to
> reasons that don't seem silly to you. :-)

Oh, I wouldn't dream of you doing that... :)

However, I do have to point out that your reason doesn't make much  
sense.

You won't use one compiler because it has bugs which also appear in  
the one you will use.

That just makes no sense to me, but that's OK.

>> Doubtless.  If I run into one as easy to trigger and as serious as  
>> that
> one, well, I'l deal with that if/when it happens.

Run almost anything but the latest 4.x versions, and you'll see worse  
than that.

I'm just hoping that the versions I'm using now don't have it.  I ran  
some mixes of code to try and flush things like that out, and so far  
the latest compiler is much better compared to the early 4.x series.

Now if they could only solve the C++ library issue somehow...

>> There are also hardware bugs in the older CPUs.  For example, the Sun
>> I sent to you has a CPU bug that will bring the machine down hard
>> from userland with a 1 instruction assembly program.
>
> !!  Where can I find out more?

http://www.hoobie.net/security/exploits/hacking/sparc_cpu.txt

There are other reports of it here and there if you look around.

Sun and Fujitsu made sure there was a workaround for Solaris, and  
supposedly fixed other more obscure bugs than this one, but have  
refused to tell anyone what they are.

>> Did you try reporting this problem to the GCC guys?  Reporting it on
>> a NetBSD list isn't likely to do much.
>
> No; last time I wrote to the gcc people, I was told I was using an
> antique compiler version, even though it was the most recently NetBSD
> release's version, with the implication that this meant nobody would
> care.  While that was offering an improvement rather than fixing a  
> bug,
> I definitely don't feel any particular compulsion to heatseek gcc
> versions just for the sake of helping *them* fix *their* bugs.

That's one of the things I hate about GNU and to a certain extent, a  
lot of open source: no discipline.  They keep releasing a constant  
stream of new code, without ever stopping long enough to stabilize  
anything.

Only a small number of people manage to do that kind of hard work,  
and they often have to fight the other 90% of the team to do it.

>> Personally, I think the industry has a dire need of a better
>> universal compiler, and one without the GPL.
>
> The industry has a dire need of non-GPL versions of lots of GPL
> stuff...but, admittedly, gcc is arguably the most urgent case.

Also, the compiler and a truly open C library would be better for  
getting start on the other stuff.

-- 
"Where some they sell their dreams for small desires."