On Sep 16 2013, at 01:15 , Magnus Ihse Bursie wrote: > On 2013-09-15 23:32, Mike Duigou wrote: >> Hello all; >> >> This is a fix for an annoying little bug. Currently ABS_TEST_OUTPUT_DIR in >> test/Makefile is defined relative to a test root of it's choosing rather >> than in terms of the TEST_OUTPUT_DIR variable. Because ABS_TEST_OUTPUT_DIR >> definition also includes necessary creation of the target directory and >> conversion to cygpath it's not feasible to just override ABS_TEST_OUTPUT_DIR >> the way it is to override TEST_OUTPUT_DIR. ie. make TEST_OUTPUT_DIR=foo >> >> This patch defines ABS_TEST_OUTPUT_DIR using TEST_OUTPUT_DIR's definition >> which makes it easier to override the output location. This fix will allow >> the changes which were part of JDK-8023491 to control the output location to >> be reattempted. >> >> http://cr.openjdk.java.net/~mduigou/JDK-8024842/0/webrev/ > > Hi Mike, > > This is similar to a fix I wrote almost a year ago, during build-infra > development, that seems to have gotten lost on the way. > > The jdk/test/Makefile uses a similar system for determining test output, > however, it names this variable for overriding ALT_OUTPUTDIR.
I prefer that name as well. At this point I have tried, though not always successfully, to avoid breaking existing users. > > Do you think this would be a good idea to get the two test Makefiles to be > more similar? That is my (incremental) goal. Eventually I hope to have little more than a stub Makefile for test in each project. I had another patch which was more aggressive about making the langtools and jdk test/Makefiles more similar back in June, JDK-8016577, but it was too aggressive at the time. I hope to revisit it eventually. > > /Magnus
