RFR[P1]: 8055744 - 8u-dev nightly solaris builds failed on 08/20

[Daniel D. Daugherty]

On 8/21/14 10:19 AM, Jesper Wilhelmsson wrote:
> Hi,
>
> On Solaris the HS_ALT_MAKE variable was not passed to vm.make when 
> creating the mapfiles which lead to mapfile-ext not being found and 
> later a linker error due to symbols declared in the extra mapfile not 
> being found.
>
> The hotspot makefiles are .. interesting .. yes.
>
> The proposed solution is to include defs.make where HS_ALT_MAKE is set 
> up into vm.make on Solaris.
>
>
> Webrev:
> http://cr.openjdk.java.net/~jwilhelm/8055744/webrev/

Unfortunately, including defs.make in make/solaris/makefiles/vm.make
isn't the "right" way to get a top-level variable down into the
HotSpot build system. I cannot remember what breaks when you do that,
but it doesn't work right in all the ways that we build HotSpot.

For the HotSpot build system, you'll want to:

- update make/solaris/Makefile and add HS_ALT_MAKE to
   the BUILDTREE_VARS list:

   BUILDTREE_VARS += HS_ALT_MAKE=$(HS_ALT_MAKE)

- update make/solaris/makefiles/buildtree.make and add
   HS_ALT_MAKE to the rule that creates flags.make:

           [ -n "$(HS_ALT_MAKE)" ] && \
             echo && echo "HS_ALT_MAKE = $(HS_ALT_MAKE)"; \

The way I usually find all the right spots is I look for
where ZIPEXE is added to the above files and follow those
examples.

Dan




>
> Bug:
> https://bugs.openjdk.java.net/browse/JDK-8055744#comment-13542110
>
> This is a P1 so if you feel comfortable with the hotspot makefiles, 
> please have a look.
>
> Thanks!
> /Jesper