Re: [RFC] Fail builds when regression check shows new errors.
On Thu, Feb 19, 2004 at 11:46:58PM -0500, Jeff Bailey wrote:
> No, because that penalizes the common case (which could be expensive on
> platforms like m68k).
I don't understand your comment, how does this penalize m68k?
If we build multiple passes the logfiles have different names (appended
$(curpass)). This addition is really for the developers who want to see
the testsuite run again if you use 'debuild -nc' after removing
check_libc.
> A rule like $(pass)_clean_tests or something like that might make sense
> where it resets the whole getup to a pre-make check state for debugging,
> though.
Now that is a big step, I don't want to *force* a recompile of all the
tests. If the tests source, or things that the test depends on have
changed, then the test will recompile itself. I just want to remove any
test output files before running the tests, by removing the output files
you force a test re-run (and only for the current pass).
c.
--- build.mk.orig 2004-02-21 12:15:45.787580520 -0500
+++ build.mk 2004-02-21 12:24:50.547764384 -0500
@@ -69,6 +69,9 @@
echo "Kernel too old, skipping tests."; \
else \
echo Testing $(curpass); \
+ echo -n 'Removing previous output from testsuite runs...' \
+ find $(DEB_BUILDDIR) -name '*.out' | xargs rm \
+ echo 'done.' \
$(MAKE) -C $(DEB_BUILDDIR) -j $(NJOBS) -k check 2>&1 | tee -a $(log_test); \
fi
touch $@
Reply to: