On Wed, Apr 27, 2016 at 07:52:41AM +0700, Duy Nguyen wrote: > > Can you (or somebody else) double check that the resulting Makefile > > gets the build dependencies and exec path right? > > > > I am seeing occasional failure from t0040 when checking out between > > master and pu, and between the branches test-parse-options do change > > behaviour which explains the breakage. > > There is a problem moving between master and next/pu though. > bin-wrappers is not regenerated after the move so it could point to > the old binaries (in the other place). Not sure how to fix that > cleanly, will think of something, maybe, in about 8 hours. This patch forces bin-wrappers regeneration every time a test program is updated. A bit wasteful, but I don't see a better option (which is also why I limit this to test programs only). A similar patch (without "t/helper/") could be applied on maint/master, otherwise bin-wrappers is still not updated when moving back from next/pu. And of course the problem will pop up again if you build old history... -- 8< -- diff --git a/Makefile b/Makefile index dd178ee..6b6a844 100644 --- a/Makefile +++ b/Makefile @@ -2196,10 +2196,13 @@ GIT-PYTHON-VARS: FORCE endif test_bindir_programs := $(patsubst %,bin-wrappers/%,$(BINDIR_PROGRAMS_NEED_X) $(BINDIR_PROGRAMS_NO_X) $(TEST_PROGRAMS_NEED_X)) +bindir_test_programs := $(patsubst %,bin-wrappers/%,$(TEST_PROGRAMS_NEED_X)) all:: $(TEST_PROGRAMS) $(test_bindir_programs) all:: $(NO_INSTALL) +$(bindir_test_programs): bin-wrappers/%: t/helper/%$X + bin-wrappers/%: wrap-for-bin.sh @mkdir -p bin-wrappers $(QUIET_GEN)sed -e '1s|#!.*/sh|#!$(SHELL_PATH_SQ)|' \ -- 8< -- -- Duy -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html