So we're already successfully using this patch, it solves our real world problem, it's a simple effective solution, but it points out architectural problems that I would want to address if I was maintaining the build system. I might waste too much time re-reading "Recursive Make Considered Harmful"
On Wed, Oct 25, 2017 at 1:09 AM, Magnus Ihse Bursie < magnus.ihse.bur...@oracle.com> wrote: > We have implicitly assumed that subsequent calls to $(wildcard) will > return the same (sorted) order of files during a single build operation. > This is not guaranteed since GNU make 3.82, but will be the case in > practice on most systems due the the filesystem being used. > > It turned out that if the order was different, FindAllModuleInfos returned > different strings which in turn caused module-deps.gmk to be re-generated > at any time during the build, causing race conditions. > > The bug analysis and patch is contributed by Martin Buchholz. > > Bug: https://bugs.openjdk.java.net/browse/JDK-8189376 > Patch inline: > diff --git a/make/common/Modules.gmk b/make/common/Modules.gmk > --- a/make/common/Modules.gmk > +++ b/make/common/Modules.gmk > @@ -249,10 +249,10 @@ > # configuration. > # Param 1 - Module to find for, set to * for finding all > FindAllModuleInfos = \ > - $(wildcard \ > + $(sort $(wildcard \ > $(foreach sub, $(SRC_SUBDIRS), \ > $(patsubst %,%/$(strip $1)/$(sub)/module-info.java, > $(TOP_SRC_DIRS))) \ > - $(patsubst %,%/$(strip $1)/module-info.java, > $(IMPORT_MODULES_SRC))) > + $(patsubst %,%/$(strip $1)/module-info.java, > $(IMPORT_MODULES_SRC)))) > > # Find module-info.java files in the specific source dir > # Param 1 - Src dir to find module-info.java files in > > /Magnus >
