Autrijus, On Saturday, March 30, 2002, at 07:20 , Autrijus Tang wrote: > On Sat, Mar 30, 2002 at 07:07:34AM +0900, Dan Kogai wrote: > > mm. In a sense, .ucm files is like .c files -- it can contain comments > and even embedded documentations (CWEB, heh), but in the end of day > it's the .so and manpages that should get installed, not the source > itself. > [snip]
I have to admit your remarks are more convincing than mine. Still shaky with qualitative approach to make a decision, I tried a quantitative approach. HOW MUCH of the space is actually taken? And here is the result on my FreeBSD box. 32743 $prefix/lib/perl5/5.7.3 9304 $prefix/lib/perl5/5.7.3/i386-freebsd/auto/Encode 8088 $prefix/lib/perl5/5.7.3/Encode Encode makes up some 53.5%! Though still smaller than mighty Emacs 21 with 42 MB on $prefix/share/emacs/21.1, This is mindbogging enough for me. 46% of wait loss of Encode itself and 25% wait loss of PERL5LIB total is way too insatiable. But this is the last remark of yours that has pushed. > I would like to have a makemaker flag that controls the installation > of ucm files. But by default they shouldn't be installed, imho. Yeah, we can always resort to Makefile.PL, with it self being a code you can do virtually any trick. I am yet to implement the trick part of MM flag, I have just relocated *.ucm to freshly-dug ucm/ directory, with Encode/encode.h stay unmoved for the later use by enc2xs And here is the result. 77 $prefix/lib/perl5/site_perl/5.7.3/i386-freebsd/Encode 9304 $prefix/lib/perl5/site_perl/5.7.3/i386-freebsd/auto/Encode With Encode directory now so clean and empty with just encode.h, I am planning to move some of the Heredocumented codes for Makefile.PL generation there so we can debug more easily. And by the time of next release, MM option will be added so you can install *.ucm if you want. These changes will be significant enough to jump version from 1.01 to 1.10. So the next one will be 1.10. Stay tuned! Dan the Encode Maintainer
