Looks great! Nice cleanup/abstraction. By "multiple" calls, you mean nested calls, not concatenated. This was not immediately clear to me. Maybe clarify that in the comment to PathList. Also, it mentions -classpath but this applies to modulepath as well.
/Magnus > 12 apr. 2016 kl. 15:44 skrev Erik Joelsson <erik.joels...@oracle.com>: > > This failed on Windows in the workaround for jdk.vm.ci modulesourcepath. Here > is a new version where I made the PathList macro safe for multiple calls. I > also added a test to verify this behavior. > > http://cr.openjdk.java.net/~erikj/8153969/webrev.top.03/ > > /Erik > >> On 2016-04-12 14:37, Erik Joelsson wrote: >> New webrev with some more cleanup in CompileJavaModules.gmk. Fixed the >> variable name and also realized that the call to MakeDir was no longer >> necessary. Starting to look pretty neat. >> >> http://cr.openjdk.java.net/~erikj/8153969/webrev.top.02/ >> >> /Erik >> >>> On 2016-04-12 11:37, Erik Joelsson wrote: >>> That is a good point, I will clean up that naming too. >>> >>> /Erik >>> >>>> On 2016-04-11 18:26, Jonathan Gibbons wrote: >>>>> On 04/11/2016 05:28 AM, Erik Joelsson wrote: >>>>> We currently have 3 places in the makefiles where the full list of source >>>>> dirs for each module is calculated. I have refactored this into macros in >>>>> make/common/Modules.gmk so that it's handled uniformly. This will also >>>>> make it simpler for other potential users of this data, like IDE project >>>>> generation. >>>>> >>>>> Bug: https://bugs.openjdk.java.net/browse/JDK-8153969 >>>>> Webrev: http://cr.openjdk.java.net/~erikj/8153969/webrev.top.01/ >>>>> >>>>> /Erik >>>> >>>> In a world where we are trying to differentiate betwen module paths and >>>> class paths, the name MODULECLASSPATH seems somewhat schizophrenic. >>>> >>>> -- Jon >
