Ilya Zakharevich wrote:
On Mon, Dec 18, 2006 at 04:02:06PM -0500, Michael G Schwern wrote:
I've got questions about the MakeMaker portion of the patch.
"doc_inst_perl" target may get confused in absense of Makefile.PL;
Under what conditions is there no Makefile.PL?
Here is how the static build is performed (in perlos2.pod=README.os2,
section "Building a binary distribution", part "7"):
After building dynamic binary distribution, we go to the CPAN build
directory (with huge limit, so it contains all the built modules
expanded), and create a tiny Makefile.PL in the root:
use ExtUtils::MakeMaker;
WriteMakefile NAME => 'dummy';
Running this Makefile.PL with a statically linked perl will create
static versions of all the distributions in the subdirectories.
But some of the distributions are broken, and contain no
Makefile.PL.
This confuses a particular branch of MakeMaker (do not know more
details, but this chunk fixed it ;-).
If there's no Makefile.PL, where did the Makefile come from? I know CPAN.pm
will generate one if none exists, does it have to do with that?
enable parallel build of subdirs (useful due to the way static
perl binary distribution is created).
Could you explain this further? The code does not appear to be
changing the building of subdirs but the testing of them. Is the
idea to add a new subdirs-test target so that parallel make has a
target to parallelize thus allowing the top tests and subdir test to
run simultaneously?
I assume there are many (100s?) of subdirs. This allows testing them
in parallel. Since many tests include sleep()ing, this may speed up
the development cycle a lot.
How is this related to static builds?
See above. When I do a static build, there is much more than a
hundred of subdirectories...
Ok, I'll buy that.
