Hello Stefano, * Stefano Lattarini wrote on Tue, Apr 12, 2011 at 03:26:33PM CEST: > I've pushed the attached patch to yacc-work. The new testcase > should correctly capture and expose the regression, as it passes > with automake 1.11 and fails with developement (maint, master > and yacc-work) versions of automake.
Thanks for all your testing work. However, I think it is not ideal if we add tests for regressions on branches other than where the regressions occurred. When we fix the regression, we want the test in place right where the fix is needed, i.e., ideally on a new branch off of the offending commit just like we'd do with a fix. Am I gathering correctly that the src/foo.c: src/$(dirstamp) rule lets GNU make build the thing in the wrong directory due to the target lookup rules? Have you analyzed this failure, and if yes, could you please explain it? Thanks, Ralf > Subject: [PATCH] coverage: test for automake bug#8485 (known regression) > > * tests/yacc-dist-nobuild-subdir.test: New test. > * tests/Makefile.am (TESTS, XFAIL_TESTS): Update. > index 0000000..b6811d7 > --- /dev/null > +++ b/tests/yacc-dist-nobuild-subdir.test > + > +# Check that VPATH builds and "make distcheck" works with packages > +# using yacc and the automake 'subdir-objects' option. > +# Exposes automake bug#8485. [...]
