Dropping 32-bit support (was Branches) (original) (raw)
David Holmes david.holmes at oracle.com
Tue Feb 21 17:15:16 PST 2012
- Previous message: Dropping 32-bit support (was Branches)
- Next message: Dropping 32-bit support (was Branches)
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
On 22/02/2012 9:49 AM, Phil Race wrote:
I was half suggesting that, half wondering. I suspect that most of the queries on the value of that flag are inside some "if " test so we may only need to make the OS X build path understand the combined setting for those cases, minimising the disruption.
Some of them are "generic" in hotspot (eg in the else of a !windows check or a !zero check). Didn't check JDK side.
David
-phil.
On 2/21/2012 3:32 PM, Mike Swingler wrote: Would it be feasible to overload ARCHDATAMODEL to take a string like "32+64", "32/64", or "Universal" (even though that's an ambiguous misnomer)?
On Feb 21, 2012, at 3:31 PM, Phil Race wrote:
'ARCHDATAMODEL' has historically been used to select a build as 32 or 64 bit. So 'ARCHDATAMODEL=32' ought to be able to select a 32 bit only build without too many build changes. Making that support 32+64 as well may be appropriate for OS X builds.
-phil.
On 2/21/2012 3:19 PM, Mike Swingler wrote: Awesome, thanks much James. Is there a pre-existing flag or naming convention for architecture-related build flags? Obviously we want to align with existing precedent, and get the HotSpot and JDK sides to use the same flag. :-) On Feb 21, 2012, at 2:58 PM, James Melvin wrote: Yes. I can file a bug and look into this, Mike. - Jim
On 2/21/12 5:55 PM, Mike Swingler wrote: On Feb 21, 2012, at 2:45 PM, James Melvin wrote: One caveat... For the JVM, we've preserved 32/64-bit universal builds. Currently, the JVM universal build only includes 64-bit support. Additionally including 32-bit requires 3 Makefile uncomments. However, there may likely be additional work on the JDK side to fully support the same. That's good to know, but that should really be keyed off of a build flag (which can default to 64-bit only). Should we file a CR through the Oracle bug reporter to get the process started to change this for HotSpot? Is there someone who can help restore the 32/64-bit build-ability for the JDK side, or at least direct us as to where we can start filing a CR? Curious, Mike Swingler Apple Inc.
- Previous message: Dropping 32-bit support (was Branches)
- Next message: Dropping 32-bit support (was Branches)
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]