On Dec 7, 2010, at 9:54 AM, Eric MSP Veith wrote:
> Am Dienstag, 7. Dezember 2010, 04:43:10 schrieb Jeff Johnson:
>> Talk to Per Oyvind if its the rpm-5_3 you want to change. rpm-5_4 is the
>> same as HEAD, including your changes from this weekend.
>
> Per, do I have your permission? Or, more appropriately, do you want to merge
> the corresponding changes from HEAD? Basically, it's about the following
> patch:
>
> ---%<---
> RCS file: /v/rpm/cvs/rpm/configure.ac,v
> retrieving revision 2.470
> diff -p -u -w -r2.470 configure.ac
> --- configure.ac 2 Nov 2010 15:57:10 -0000 2.470
> +++ configure.ac 7 Dec 2010 14:52:36 -0000
> @@ -1517,8 +1517,17 @@ RPM_CHECK_LIB(
> [Ruby], [ruby],
> [ruby], [ruby_init], [ruby.h],
> [no,external:none], [],
> - [ AC_DEFINE(WITH_RUBYEMBED, 1, [Define to 1 if you want embedded Ruby])
> + [
> + AC_DEFINE(WITH_RUBYEMBED, 1, [Define to 1 if you want embedded Ruby])
> WITH_RUBY_SUBDIR=ruby
> +
> + WITH_RUBY_CPPFLAGS=["-I`ruby -rrbconfig -e \
> + 'puts RbConfig::expand(RbConfig::CONFIG[\"rubyhdrdir\"])'` \
> + -I`ruby -rrbconfig -e \
> + 'puts
> RbConfig::expand(\"#{RbConfig::CONFIG[\"rubyhdrdir\"]}/$(arch)\")'`"
> + ]
> + WITH_RUBY_VENDORARCHDIR=[`ruby -rrbconfig -e \
> + 'puts RbConfig::expand(RbConfig::CONFIG["vendorarchdir"])'`]
> ], [])
> AC_SUBST(WITH_RUBY_CPPFLAGS)
> AC_SUBST(WITH_RUBY_SUBDIR)
>
The main problem here is self-hosting: one cannot build with
ruby unless ruby is functional. That logic has some corner
cases with cross-packaging when host != target,and forces
an additional build dependency on /usr/bin/ruby that
is not easily tracked with AutoFu.
(aside)
There is scripts/ruby-config, and it isn't that hard to cruft
up either ruby.pc or libruby-X.Y.la ... all preferred to
invoking ruby imho.
There's the further issue of what happens if you have multiple
versions of ruby installed, where ruby-X.Y or /opt/local/bin/ruby
must be invoked. Its not clear (nor necessary) that /usr/bin/ruby
determines the library to link against.
A better approach is declartive, like adding ruby-config or
ruby.pc or libruby.la, each of those hintings for build flags
is more narrowly targeted, and easier to integrate in AutoFu.
> --->%---
>
>> Embedded, not bindings, is the issue. There's no way to do 1.9.x
>> embedding without a thread co-routine still afaik.
>
> That's true, but the problem lies mainly in querying the correct
> inlcude/libraries paths. I.e., on my system, simply including <ruby/ruby.h>
> doesn't work. One needs /usr/include/ruby/$ARCH as an additional per-
> architecture include path. So it's neccesssary for both bindings and embedded.
>
> Meh.
>
> The flaw that is hacksolved by the coroutine solution doesn't become apparent
> until runtime AFAIK. And its IMHO currently the better alternative as it
> allows multiple interpreter instances and proper memory pooling, doesn't it?
>
embedded is all about run-time yes.
Meanwhile the problem I reported _IS_ with the configgery you are suggesting
moving to rpm-5_3, I saw that flaw on rpm-5_4 (which also means on HEAD).
WORKSFORME on rpm-5_3 (but I likely have ruby-config twisted into my machine
somewhere).
73 de Jeff
______________________________________________________________________
RPM Package Manager http://rpm5.org
Developer Communication List rpm-devel@rpm5.org
