We should add a new high-level boolean configure option: --enable-java-debug-symbols/--disable-java-debug-symbols. This avoids issues such as JDK-8145115.

From the review thread of JDK-8145115:
On Fri, 2015-12-11 at 14:10 +0100, Magnus Ihse Bursie wrote:
> On 2015-12-10 16:02, Severin Gehwolf wrote:
> > Yes, pretty please :) This is a long standing issue (not just java
> > class debug info, but debug info in general). See
> > JDK-8036003 with review thread:
> > http://mail.openjdk.java.net/pipermail/build-dev/2015-December/016155.html
> >
> > Once JDK-8036003 is fixed we'd need something similar for Java classes
> > and we can get rid of invoking make with build internal variables.
>
> I agree. Yasumasa Suenaga had a pretty good implementation in
> http://cr.openjdk.java.net/~ysuenaga/JDK-8036003/webrev.04/.
>
> I think it could form the basis of getting a proper java debug symbol
> support flag to configure. My main issue with that is that it must also
> work with the case "--enable-debug --disable-java-debug-symbols", to
> produce a build without java debug symbols. This means that this logic
> need to interact with the logic in flags.m4:
> case $DEBUG_LEVEL in
> fastdebug | slowdebug )
> CFLAGS_JDK="$CFLAGS_JDK $CFLAGS_DEBUG_SYMBOLS $CFLAGS_DEBUG_OPTIONS"
> CXXFLAGS_JDK="$CXXFLAGS_JDK $CXXFLAGS_DEBUG_SYMBOLS
> $CXXFLAGS_DEBUG_OPTIONS"
> JAVAC_FLAGS="$JAVAC_FLAGS -g"
>
> Without digging a bit deeper in the source, I can't say the best way to
> achieve this. Perhaps the configure flag check should set a variable
> like JAVA_DEBUG_SYMBOLS:=true/false/default that can be checked in
> flags.m4? Or we should defer the modification of JAVAC_FLAGS.
>
> /Magnus