Is it your intent to become a official/unofficial distributor of
ming_win32 distro?
Or is it a one-time deal?
I plan on continuing to use it for personal use (as long as it's faster,
I'm gonna use it myself, and offer it as an unofficial distro). I also
am in touch with the OCI guys and they may adopt it and/or offer it as a
branch off main.
In terms of 'will it be supported' thankfully (you'll think this is a
cop out but)--the extensions compiled with VC6 and mingw are binary
compatible, so in reality you could just drop the mingw
ruby.exe+ruby18.dll over the existing ones, rename your i386-win32
folder i386-mingw and it would work, (I think--these things get a little
complicated at times). So...it's about the same.
The OCI guys haven't decided if they're going in the direction of VC8 or
mingw. This is the mingw branch.
you say that that it seems faster .... wish full thinking or actual
benchmarking?
It is typically just as fast (I/O bound) or up to 1/3 faster or more
(CPU bound).
Here's one I'll do to demo
old one (VC6)
/cygdrive/c/old_ruby_msvc/bin/ruby -e "1000.times {'a'*20000}"
real 0m1.468s
new one (mingw)
ruby -e "1000.times {'a'*20000}"
real 0m0.543s
(consistently gets these speeds). From my own experience of developing
rails apps on it, it feels faster. In all honesty, rails finally feels
like it WORKS with the mingw build, unlike feeling broken the old
way--in terms of slugishness. Some unofficial speed comparisons are
also given here
http://pastie.caboo.se/pastes/95210
I'd imagine it would compile fine with mingw but haven't taken the
chance to do so and compare, as I don't use it.
There are probably a few bugs still which I'm enlisting help to hunt
down. (like -e is broken if you have quotes within quotes), but I have
been using it for two weeks with minimal probs and nicer speed. Once
they're ironed out it should be a real distro (albeit not one click). I
believe for example it is still lacking "readline.dll" in the bin
directory, but we're getting there.
-Roger
