jdk9 windows build failure

Magnus Ihse Bursie magnus.ihse.bursie at oracle.com
Thu Apr 16 09:02:37 UTC 2015


On 2015-04-15 01:44, David Katleman wrote:
> We went through this exercise in January.
>
> RE intentionally uses what is the documented minimum gnumake (3.81), 
> to verify that builds work with the oldest supported gnumake.
>
>> On 2015/1/31 4:01, David Katleman wrote:
>>
>> You need to revert your 4.0 change make to JDK9 builds.   Until dev 
>> upgrades the minimum gnumake used in the JDK 9 builds to 4.0, RE must 
>> continue to build with the current official minimum, which is 3.81, 
>> to catch that the official minimum still works, as in JDK-8071651 
>> <https://bugs.openjdk.java.net/browse/JDK-8071651>.
>>
>> I've opened up a high priority enhancement request to get the 
>> official gnumake updated, JDK-8072023 
>> <https://bugs.openjdk.java.net/browse/JDK-8072023>.
>
> Until we update gnumake for windows (JDK-8072023 
> <https://bugs.openjdk.java.net/browse/JDK-8072023>), issues like what 
> Phil found with gnumake 3.81 should be fixed.

I'm not really comfortable with letting go of support for GNU Make 3.81. 
It's been around for a long time and was the latest stable release for a 
long time when GNU Make didn't make much progress. While Windows is 
special, since cygwin more easily pushes newer versions on users, if 
this is indeed a problem with make 3.81 maybe we should try to go to the 
bottom with it.

What makes java.xml.bind so special that make 3.81 should barf on it, 
but no other modules?

/Magnus



More information about the build-dev mailing list