On 2020-05-01 19:33, Erik Joelsson wrote:
After the sjavac cleanup in JDK-8244036 (and the subsequent fix of the
javac server usage in JDK-8244210), two more build failures have been
noted. The bootcycle build fails and the test-make target fails the
tests of SetupJavaCompilation.
The bootcycle build fails because a space has crept into the variable
SJAVAC_SERVER_DIR when it's overridden in bootcycle-spec.gmk.in This
space splits the sjavac server argument.
The tests for make just need to be updated to use the new
TARGET_RELEASE arg instead of the old SetupJavaCompiler construct.
Bug: https://bugs.openjdk.java.net/browse/JDK-8244247
Webrev: http://cr.openjdk.java.net/~erikj/8244247/webrev.01/index.html
The change looks good. However, it makes me realize that the variable
should really have been renamed JAVAC_SERVER_DIR. Do you think you could
fix that in the same bug? Otherwise I'll do it separately.
I also just wanted to note that technically, the assignment in
bootcycle-spec.gmk.in has not changed -- the extra space has always been
there. But due to how it was processed, that additional space is no
longer stripped when the variable is read. It was not explicitly
stripped before either, but it went a second round through the makefile
parsing by being used in SetupJavaCompiler as "SERVER_DIR :=
$(SJAVAC_SERVER_DIR)", which effectively stripped it.
/Magnus
/Erik