RFR: JDK-8173107 Fix autoconf/spec.gmk mismatches

[David Holmes]

On 20/01/2017 8:22 PM, Magnus Ihse Bursie wrote:
> On 2017-01-20 10:30, David Holmes wrote:
>> Hi Magnus,
>>
>> On 20/01/2017 6:58 PM, Magnus Ihse Bursie wrote:
>>> Over the course of jdk9 development, the variables in the configure
>>> script exported using AC_SUBST, the replacement patterns in spec.gmk.in
>>> and the usage of spec.gmk in the makefiles have diverged.
>>>
>>> Fixes:
>>>
>>> * Remove AC_SUBST for variables not exported or not used in make files.
>>> (BOOT_JDK_BITS, ZERO_ARCHFLAG, ZERO_ARCHDEF, INCLUDE_DTRACE,
>>
>> I'm not sure what purpose the whole --enable-dtrace section serves if
>> we don't export or use INCLUDE_DTRACE ??
>
> Ah, but we do use INCLUDE_DTRACE, not just in the make files. :-)
>
> INCLUDE_DTRACE is a local variable in hotspot.m4, which is used to
> enable the "dtrace" JVM feature. (The list of JVM features is how we
> communicate this from autoconf to the makefiles). The

Ah! Yes I missed that.

> AC_SUBST(INCLUDE_DTRACE) should have been removed with the new hotspot
> build, at the same time as @INCLUDE_DTRACE@ in spec.gmk.in was removed.

Understood.

The changes seem okay but the proof here is in the building. Not sure 
about things that affect s390 and zero for example. Hopefully those folk 
can chime in on that.

Thanks,
David

> /Magnus
>
>>
>> Thanks,
>> David
>>
>>> OPENJDK_*_OS_BUNDLE, OPENJDK_*_CPU_BUNDLE, LP64, ENABLE_JFR)
>>> * Include missing exports in spec.gmk.in (JDK_ARCH_ABI_PROP_NAME)
>>> * Make sure we always explicitly use AC_SUBST for exported variables
>>> (PNG_CFLAGS, PNG_LIBS, LCMS_CFLAGS, LCMS_LIBS)
>>> * Only use the pattern AC_SUBST(VARIABLE), to make sure the VARIABLE is
>>> accessible in configure as well as make.
>>>
>>> Bug: https://bugs.openjdk.java.net/browse/JDK-8173107
>>> WebRev:
>>> http://cr.openjdk.java.net/~ihse/JDK-8173107-fix-spec-gmk-mismatches/webrev.01
>>>
>>>
>>>
>>> /Magnus
>>>
>