cvs commit: src/lib/libc/i386/net htonl.S ntohl.S

[Scott Long]

Ruslan Ermilov wrote:
> On Mon, Oct 18, 2004 at 11:42:34AM -0700, David O'Brien wrote:
> 
>>On Mon, Oct 18, 2004 at 08:31:18PM +0200, Erik Trulsson wrote:
>>
>>>On Mon, Oct 18, 2004 at 10:45:11AM -0700, David O'Brien wrote:
>>>
>>>>Yes.  You missed the discussion in freebsd-current on this?
>>
>>..
>>
>>>I have certainly not seen any message stating that on -current.
>>
>> To: freebsd-current at freebsd.org
>> Subject: [PATCH] optimizing X now that we don't support i386 CPU's on 6-CURRENT
>> Message-ID: <20041016174925.GA96809 at dragon.nuxi.com>
>>
> Well, it's less than three days old, and you made a major
> decision with this commit that the world no longer supports
> i386 by default.  I wish you discussed in on arch@ first.

This has been discussed for years.  It should not be a surpise.  Frankly
I'm thrilled that David at least made it conditional rather than just
culling the support entirely as has been threatened so often in the
past.

> 
> I re-read this thread, and it seems that you think that since
> nobody reported a success running on i386 then it's time to
> kill I386 support in our world by default?

i386 hasn't been supported in the default configuration for years. 
Whether or not someone got it running with 5.2.1 doesn't change this.
The knobs are there to (theoretically) turn it on.  So as long as those
knobs are consistent and documented, nothing is lost.

> 
> I can also read a message from John Nielsen in this same
> thread saying he was able to install 5.2 on 386DX, so perhaps
> you owe a dinner to Xin next time you're in Bejing.  ;)
> 
> 
>>>I know that the GENERIC kernel does not work on a i386, so that you
>>>need a custom-built kernel to install on a real i386.
>>>I also know that support for FPU-less systems has been dropped, which
>>>means that most 386-based systems (those without an '387) (as well as
>>>systems with a 486sx w/o a 487sx) will not run RELENG_5 or -CURRENT.
>>>
>>>Neither of these means that a real 80386+80387 is unsupported, just
>>>that you need to build a new kernel first.
>>
>>I see nothing wrong with having to build a new userland before it will
>>run on I386 since we already have to do that with the kernel.
>>
> 
> Having libc optimized for >386 by default is good (IMO), but what
> you've committed can't be used easily to compile for i386.  Perhaps
> you will adopt/apply my suggestions, and then we will fix the issue
> for i386 users.  I agree with Xin that this deserves a HEADS UP
> since default world will no longer support i386 hardware.
> 
> 
> Cheers,

Guys, just decide on the name of the knob and be done with it.  Please!
This was settled years ago.  The 80386 isn't making an unexpected
comeback here that warrants a lot of fighting.

Scott