jdk HEAD fails to build?

[Severin Gehwolf]

On Tue, 2018-11-13 at 09:44 -0800, Erik Joelsson wrote:
> I'm able to reproduce this when setting LOG=debug.

Thanks for checking, Erik.

Cheers,
Severin

> /Erik
> 
> On 2018-11-13 09:17, Severin Gehwolf wrote:
> > Hi Erik,
> > 
> > On Tue, 2018-11-13 at 09:06 -0800, Erik Joelsson wrote:
> > > Have you tried after pulling in
> > > https://bugs.openjdk.java.net/browse/JDK-8213338?
> > 
> > Yes, I have that. I've got it narrowed down to:
> > https://bugs.openjdk.java.net/browse/JDK-8213736?focusedCommentId=14222442&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-14222442
> > 
> > I have no idea why 'LOG=debug' causes this:
> > 
> > $(shell $(MAKE) -s --no-print-directory $(MAKE_ARGS) \
> >      SPEC=$(SPEC) -f $(TOPDIR)/test/make/TestMake.gmk print-targets)
> > 
> > to also contain:
> > gmake[1]: Leaving directory '<top-dir>'
> > 
> > It seems debugging output. All I know is if I use this patch it starts
> > to build again with LOG=debug:
> > 
> > diff --git a/make/common/FindTests.gmk b/make/common/FindTests.gmk
> > --- a/make/common/FindTests.gmk
> > +++ b/make/common/FindTests.gmk
> > @@ -72,9 +72,7 @@
> >   # Add Gtest
> >   ALL_NAMED_TESTS += gtest
> >   
> > -# Find make test targets
> > -MAKE_TEST_TARGETS := $(shell $(MAKE) -s --no-print-directory $(MAKE_ARGS) \
> > -    SPEC=$(SPEC) -f $(TOPDIR)/test/make/TestMake.gmk print-targets)
> > +MAKE_TEST_TARGETS := make-base java-compilation copy-files idea compile-commands
> >   
> >   ALL_NAMED_TESTS += $(addprefix make-, $(MAKE_TEST_TARGETS))
> > 
> > Please help me get out of this rabbit-hole :-P
> > 
> > Thanks,
> > Severin
> > 
> > > /Erik
> > > 
> > > On 2018-11-12 03:09, Severin Gehwolf wrote:
> > > > On Mon, 2018-11-12 at 10:52 +0000, Andrew Haley wrote:
> > > > > On 11/12/2018 10:41 AM, Severin Gehwolf wrote:
> > > > > > make/Main.gmk:1056: *** target pattern contains no '%'. Stop.
> > > > > > 
> > > > > > But I'm not sure that error message is very telling. It might
> > > > > > be a red
> > > > > > herring.
> > > > > > 
> > > > > > I'll keep looking.
> > > > > 
> > > > > Does "hg status" reveal anything that shouldn't be there?
> > > > 
> > > > Thanks. It could be. If it is, it's some file which isn't tracked
> > > > and
> > > > causes issues. Having said that, I doubt it's that.
> > > > 
> > > > This works:
> > > > 
> > > > $ make JAVAC_FLAGS=-g DISABLE_INTREE_EC=true images
> > > > 
> > > > This doesn't:
> > > > 
> > > > $ make JAVAC_FLAGS=-g DISABLE_INTREE_EC=true LOG=debug images
> > > > 
> > > > So in summary it's:
> > > > - related to JDK-8210958
> > > > - possibly related to the toolchain on Fedora 28
> > > > - related to debug logging of the build:
> > > >     LOG=debug or LOG=trace fail. Default LOG doesn't.
> > > > 
> > > > I'm collecting info as I know more in:
> > > > https://bugs.openjdk.java.net/browse/JDK-8213736
> > > > 
> > > > Thanks,
> > > > Severin
> > > >