If this file can never really have execute permission in the repositories, then encouraging people to put execute permission on it creates a risk that they might try to commit that change, and then their changeset would get rejected. A big annoyance.
So why not just always run $(SH) and skip the 'if -x' test? Isn't that simplier? And will always work? -kto On Apr 23, 2012, at 3:54 AM, David Holmes wrote: > On 21/04/2012 12:14 AM, Magnus Ihse Bursie wrote: >> In the build-infra hg repo, we allowed execute permission to be set on >> files, including common/bin/logger.sh, which is called by the Makefile. >> >> The official hg repos strip this bit, so when integrating to the build >> forest, this was lost, causing the Makefile to fail since it can't >> execute logger.sh. >> >> This fix looks if the x permission is there, otherwise it calls >> logger.sh using the shell. >> >> Webrev: >> http://cr.openjdk.java.net/~ihse/logger.sh-missing-x-permission/webrev.00/ > > You could save yourself a bit of duplication using: > > BUILD_LOG_WRAPPER='$(SRC_ROOT)/common/bin/logger.sh $(BUILD_LOG)' > if test ! -x $SRC_ROOT/common/bin/logger.sh ; then > BUILD_LOG_WRAPPER='$(SH) $(BUILD_LOG_WRAPPER)' > fi > > or even factor out > > LOGGER_SH=$(SRC_ROOT)/common/bin/logger.sh > > David > ----- > >> /Magnus >>
