On Fri, Jun 07, 2002 at 04:12:28PM -0400, Michael G Schwern wrote: > On Thu, Jun 06, 2002 at 02:58:21PM -0700, David Carmean wrote:
[snip] > > My goal is to get everything tied together and in the top-level > > ..../blib so that I can develop and test subsets then the full app. > > > > The only way I've figured out to do this is to create fake > > packages at the two intervening directory levels: > > > > FOO/FOO.pm > > FOO/DDNS/DDNS.pm > > MakeMaker will only recurse one subdirectory deep, so it won't make the jump > from ./Makefile.PL to FOO/DDNS/Exceptions/Makefile.PL on it's own. Whether > or not this is the right thing, I dunno, but the work around is to simply > place a Makefile.PL in each intervening subdirectory (so one in FOO/ and one > in FOO/DDNS/) which calls ExtUtils::MakeMaker::WriteEmptyMakefile(). That > way, there will be a Makefile in each subdirectory it has to walk down > through. In fact I'm seeing that once it sees the WriteEmptyMakefile(), it seems to stop recursing at all. And even if it continued, the resulting empty makefiles in the intervening levels prevent a top-level make from recursing down into the package directories. However... I think I might have found the answer: if &ExtUtils::MakeMaker::init_dirscan doesn't find any .pm, .pl, t/*, etc files, then basically all the normal targets will just be built to recurs into subdirectories where other Makefile.PL were found. In other words: I made things worse by using h2xs to create the Makefile.PL for the two intervening levels; the resulting stub ..pm and test.pl files caused a lot of needless work. The solution seems to be just to hand-create a minimal Makefile.PL which calls WriteMakefile() with no arguments.
