Re: [PATCH 0/2] Tests initialization: put default definition of AC_CONFIG_AUX_DIR in the pre-populated configure.in.
On Monday 13 December 2010, Ralf Wildenhues wrote: * Stefano Lattarini wrote on Sat, Dec 11, 2010 at 02:52:18PM CET: This is a reproposal of an older patch of mine: http://lists.gnu.org/archive/html/automake-patches/2010-05/msg8.html http://lists.gnu.org/archive/html/automake-patches/2010-05/msg9.html I see that some tests explicitly put a call like: AC_CONFIG_AUX_DIR([.]) in the generated configure.in, to prevent automake from looking into `..' and `../..' when trying to copy required files. I think that such a provision should be made unconditional, as it could help to reduce undefined/unexpected behaviours. Obviously, we should leave a way for the individual testcases to easily override this default (and some existing tests already require such an overriding). I'm undecided on this one. On the one hand, the safety increase is a plus, but on the other hand, the tests become less readable, if only because it is less obvious what is going on. If you don't remember, you have to start reading the generated configure.in file after rerunning the test with keep_testdirs=:. Hmm. There is precedent with the 'parallel_tests' settings, Speaking against my own interest (i.e. seeing the patch applied), I must say that the 'parallel_tests' setting was more compelled, since it was needed for having generated tests that are just a thin layer around their sister tests. but even that looks unobvious to me, just like those actions triggered by a 'required' setting other than is this test run or not. Maybe we could solve this problem with a new `write_configure' subroutine in tests/defs, so that we could end up having e.g.: ./defs || Exit 1 ... write_configure --auxdir=foo configure.in 'END' AC_OUTPUT END instead of: config_auxdir=NONE ./defs || Exit 1 ... cat configure.in END AC_OUTPUT END I like this idea because, as a general rule, I think that explicit is better than implicit. Also, the `write_configure' subroutine might be easily extended in the future, to provide additional goodies. WDYT? What's the unexpected behavior we have to expect here? (Sorry for the pun.) Then there's the spelling different AUX_DIR vs auxdir. Cheers, Ralf Regards, Stefano
Re: [PATCH 0/2] Tests initialization: put default definition of AC_CONFIG_AUX_DIR in the pre-populated configure.in.
* Stefano Lattarini wrote on Mon, Dec 13, 2010 at 09:42:08PM CET: On Monday 13 December 2010, Ralf Wildenhues wrote: I'm undecided on this one. On the one hand, the safety increase is a plus, but on the other hand, the tests become less readable, if only because it is less obvious what is going on. If you don't remember, you have to start reading the generated configure.in file after rerunning the test with keep_testdirs=:. Hmm. There is precedent with the 'parallel_tests' settings, Speaking against my own interest (i.e. seeing the patch applied), I must say that the 'parallel_tests' setting was more compelled, since it was needed for having generated tests that are just a thin layer around their sister tests. but even that looks unobvious to me, just like those actions triggered by a 'required' setting other than is this test run or not. Maybe we could solve this problem with a new `write_configure' subroutine in tests/defs, so that we could end up having e.g.: That's better because of this: I like this idea because, as a general rule, I think that explicit is better than implicit. Also, the `write_configure' subroutine might be easily extended in the future, to provide additional goodies. WDYT? but it still has the same problems with respect to the test being obvious and copy-pasteable. Again: what is the risk here that this change is trying to avoid (which presumably was the reason for it in the first place)? Thanks, Ralf
[PATCH 0/2] Tests initialization: put default definition of AC_CONFIG_AUX_DIR in the pre-populated configure.in.
This is a reproposal of an older patch of mine: http://lists.gnu.org/archive/html/automake-patches/2010-05/msg8.html http://lists.gnu.org/archive/html/automake-patches/2010-05/msg9.html -*-*- Hello automakers. I see that some tests explicitly put a call like: AC_CONFIG_AUX_DIR([.]) in the generated configure.in, to prevent automake from looking into `..' and `../..' when trying to copy required files. I think that such a provision should be made unconditional, as it could help to reduce undefined/unexpected behaviours. Obviously, we should leave a way for the individual testcases to easily override this default (and some existing tests already require such an overriding). The first patch in this series implements this idea. The second one extends the tests on `AC_CONFIG_AUX_DIR' itself. OK to apply the patches to a temporary branch off of maint, and merge to the tests-init branch? I'll wait the customary 72 hours (until Tuesday evening) before pushing. Regards, Stefano
