On Tuesday 26 August 2008 03:14:50 am Ludovic Courtès wrote: > Hi, > > Christian Grothoff <[EMAIL PROTECTED]> writes: > > Please read the README file with particular emphasis on the wording > > around "GNUNET_PREFIX". This should allow you to solve part of your > > problem. > > Sorry for not seeing this!
;-) > That said, it's quite unusual and inconvenient to require one to install > the package to be able to test it. It's inconvenient because one may > notice that the new installation is "broken" once the previous one has > been overwritten. Technically yes, but I expect that "make check" is used by devs to test new code and by packagers/porters to test ports; I don't think them finding out that something does not work after doing the install is a problem in practice. > Adding the right `TEST_ENVIRONMENTS' in makefiles as I suggested would > fix it. What would you think of that? I don't think your particular change scales -- for some of the more complex tests the "TEST_ENVIRONMENTS" definition would end up being longer than the rest of Makefile.am, and it is generally non-obvious what set of directories is required. We could have one global list and then include that everywhere -- if you want to write such a patch (against Makefile.am), I'll strongly consider adding it ;-). Best, Christian _______________________________________________ GNUnet-developers mailing list [email protected] http://lists.gnu.org/mailman/listinfo/gnunet-developers
