David Golden wrote:
Comments from people who knows MB well is much appreciated. I'll try to
debug it myself, but I'm not sure I can find a source of problem quickly.
My best guess is that it's Pod::Html trying to build an index for
resolving L<> tags. C.f.
https://rt.cpan.org/Ticket/Display.html?id=68651
Yeeeees, this is it. I remember similar problem with Pod::Html when installing
PPM packages on ActivePerl.
It was fixed by removing html directory (or not installing it in the first place), then ActiveState HTML generator
wouldn't run Pod::Html.
I think that it would be a good idea to add similar check to Module::Build. It seems it runs Pod::Html even if perl/html
is absent.
My other observation on why things are slow is Makefile.PL specific --
every system operation is emulated with Perl. So a simple "copy" mean
spawning another Perl process and process spawns are very slow on
Windows. So if the Pod::Html problem were fixed, then M::B should be
faster than EU::MM by not having to spawn for every system operation.
AFAIK, these commands are running very rarely. When running "nmake test", ExtUtils::Install is used to copy files to
blib, and it uses File::Copy. I'm not even sure ExtUtils::Command commands are actually running on any stage of testing.
Maybe having MSYS's cp.exe may help in this case, I didn't look at it closely.
--
Serguei Trouchelle
