hg: build-infra/jdk8/hotspot: Hotspot uses the names i386, i486, i586 and i686 in

[Kelly O'Hair]

On Mar 25, 2012, at 10:37 PM, David Holmes wrote:

> On 24/03/2012 1:31 AM, fredrik.ohrstrom at oracle.com wrote:
>> Changeset: e78b8a52fe13
>> Author:    ohrstrom
>> Date:      2012-03-23 16:33 +0100
>> URL:       http://hg.openjdk.java.net/build-infra/jdk8/hotspot/rev/e78b8a52fe13
>> 
>> Hotspot uses the names i386,i486,i586 and i686 in
>> all sorts of places! Arggg, time for some serious clenup.
>> 
>> ! make/linux/makefiles/defs.make
> 
> Hmmmm. AFAIK the way it works now on linux is that if "uname -m" gives i?86 then ARCH is set to i686. Allowing an incoming ARCH==i586 would seem to be complicating things.
> 
> But yes it is quite insane now. For 32-bit x86 we have:
> 
> LIBARCH=i386  (used for: jre/lib/i386)
> BUILDARCH=i486 (used eg for: <outputdir>/hotspot/outputdir/Linux_i486_compiler1)
> PLATFORM=<os>_i586  (used for runtime images and bundles)
> ARCH=i686
> 
> And
> 
> SRCARCH=x86

Plus ARCHPROP  which is the value of the "os.arch" java property, which is NOT the same for all os.name's. :^(

It's a bit of a mess in my opinion. :^(

-kto

> 
> and in fact it is even more complex as solaris, linux and windows differ in the variables they use.
> 
> Don't know how this came about ... it's not like we use UltraI, UltraII, UltraIII etc for sparc. Probably something legacy from the original Windows RI.
> 
> David
> -----