Download raw body.
umask assumptions in regressions
On Thu, Apr 17, 2025 at 07:51:27PM +0100, Thomas Adam wrote: > On Fri, Mar 28, 2025 at 09:14:52PM +0000, Kyle Ackerman wrote: > > +umask 022 > > regress_run_only="" > > The problem I have with this, is the umask setting is external to the > environment which you're invoking "make tests" from. As such, if you know > thwe umask is the cause of this, simply do: > > (umask 022; make tests) > > But I'm not keen on setting an explicit umask, regardless of how innocent a > change this could be. AFAIK, "make" runs Makefile targets in a subshell. So the external umask would only be modified if we're running test scripts in a shell directly, rather than running them via make. Correct? Couldn't the -portable Makefile.am set the umask? diff /home/stsp/src/got-portable path + /home/stsp/src/got-portable commit - 91c1648f8258a8d72a8111818f85d79e2e48fbfe blob - b7158ca393f71349db4b0db82e8f3a664f51adda file + Makefile.am --- Makefile.am +++ Makefile.am @@ -54,6 +54,7 @@ tests: $(TEST_TARGETS) regress-cmdline: (export PLATFORM=@PLATFORM@; \ cd $(top_builddir)/regress/cmdline || exit $$?; \ + umask 022; \ ./checkout.sh -q -r "$(GOT_TEST_ROOT)"; \ ./update.sh -q -r "$(GOT_TEST_ROOT)"; \ ./status.sh -q -r "$(GOT_TEST_ROOT)"; \
umask assumptions in regressions