Hi Akim, * Akim Demaille wrote on Mon, Mar 30, 2009 at 09:50:59PM CEST: > Le 29 mars 09 à 00:04, Ralf Wildenhues a écrit : > >> One thing that one needs to look out for is, when overriding variables >> in recursive `make' instances is that non-GNU make don't override by >> default unless you use `make -e' and pass via the environment. >> Requiring our users to do so if they want to override variables is ok >> IMVHO, because then they are warned that their environment needs to be >> clean. OTOH using -e internally is not ok at all, as the users may >> not be aware, and their environment could cause subtle breakage. So >> consequently we can only transport overrides one recursion deep and >> not any further, unless each lower re-sets the variables on the make >> command line.
> I have not understood exactly what you are referring too, but why don't > we just pass TESTS=$(TESTS) explicitly? Or ask the users to add his > "precious macros" to AM_MAKEFLAGS? Thank you! Suddenly I noticed that your remark provides a really simple (maybe simplistic) and neat solution to one of the remaining open portability issues here. Here's a quick recap, for those of you not following closely. ;-) With Solaris make (and some others), the recheck code doesn't work correctly due to the issue described here: <http://www.gnu.org/software/autoconf/manual/html_node/Macros-and-Submakes.html> It applies in this case because there are (potentially) several 'make' recursions involved: recheck -> check -> check-TESTS -> $(TEST_SUITE_LOG) recheck-html -> check-html --> check -> check-TESTS -> $(TEST_SUITE_LOG) `-> $(TEST_SUITE_HTML) The effect is that, with Solaris make, `make recheck' will simply rerun all tests, instead of only failed ones. Possible solutions: - add TEST_LOGS=... overrides to all of the above rules in the chain, - add more helper variables for overrides to all of the above rules. The former has the already-mentioned drawback that it could run us into command line length limits, the latter is just a wee bit ugly. The patch below (ab)uses AM_MAKEFLAGS to transport things through more than one layer of sub-makes, without touching all the intermediate rules. It will break if the user sets AM_MAKEFLAGS herself and uses single quotes in that setting. Oh well, recheck is a new feature, so I guess this isn't a regression. ;-) Pushed to master. The testsuite addition merely ensures that the override not only arrives at the check-TESTS rule, but also at the $(TEST_SUITE_LOG) rule. Cheers, Ralf parallel-tests: fix recheck* targets for non-GNU make. * lib/am/check.am (recheck recheck-html): Override AM_MAKEFLAGS to portably transport the TEST_LOGS settings through the recursive `make' invocations. Fixes parallel-tests9.test failure with Solaris make. * tests/parallel-tests9.test: Also ensure that the test summary is correct for `recheck'. Suggestion by Akim Demaille. diff --git a/lib/am/check.am b/lib/am/check.am index 271720d..6653f7b 100644 --- a/lib/am/check.am +++ b/lib/am/check.am @@ -289,7 +289,7 @@ recheck recheck-html: case $$line in FAIL*|XPASS*) echo $$f;; esac; \ else echo $$f; fi; \ done | tr '\012\015' ' '`; \ - $(MAKE) $(AM_MAKEFLAGS) $$target TEST_LOGS="$$list" + $(MAKE) $(AM_MAKEFLAGS) $$target AM_MAKEFLAGS='$(AM_MAKEFLAGS) TEST_LOGS="'"$$list"'"' .PHONY: recheck recheck-html .MAKE: recheck recheck-html diff --git a/tests/parallel-tests9.test b/tests/parallel-tests9.test index e1f088b..a0cb585 100755 --- a/tests/parallel-tests9.test +++ b/tests/parallel-tests9.test @@ -65,6 +65,7 @@ cat stdout grep foo.test stdout && Exit 1 grep bar.test stdout || Exit 1 grep baz.test stdout || Exit 1 +grep '2 of 2.*failed' stdout # If we cannot read the log file, then redo it as well. chmod a-r foo.log
