* Peter Johansson wrote on Sun, Apr 04, 2010 at 05:05:38PM CEST: > I've just upgraded one of my projects to automake 1.11 in order to > use the parallel-tests driver. Both the manual [1] and the Ralf's > post here [2] describes how to run the test suite lazily as > ` make check RECHECK_LOGS='. My question is how it is determined > which tests are run in this case. > > It seems like if `foo' is newer than `foo.log', foo will be re-run > but is that the only dependency? I would like to fine tune these > dependencies. For instance, I have a test `yat_m4_test.sh' which > uses $(top_srcdir)/m4/yat.m4 and I want `yat_m4_test.sh' to be > re-run when I've modified `m4/yat.m4'.
Then you should add a dependency yat_m4_test.log: m4/yat.m4 and for dependencies to work, you need to use TEST_EXTENSIONS as documented. > As I see it there are two reasons why I would like to include a test > in a lazy 'make check' run. 1) Because the test itself has been > modified either directly or if the test is compiled by the compiler. > 2) Because any input file has changed, e.g., yat.m4 in my case or if > the test is testing a script/program it is desired to depend on the > script/program. In case 2) there is no need to update test itself > but there is a need to update the .log file, i.e., to re-run the > test. Exactly. Cheers, Ralf
