On 4/19/07, Nobuyoshi Nakada <nobu / ruby-lang.org> wrote:
> Hi,
>
> At Thu, 19 Apr 2007 22:30:47 +0900,
> Michal Suchanek wrote in [ruby-core:10970]:
> > This would break the previous fix so I did not even try to apply this ^
>
> Sorry, mistaken.
>
> > This part looks good and it allows building ruby as long as the binary
> > extension dir is different. However, after changing config.guess to
> > report the same platform as the previous build I can no longer build
> > ruby.
>
> Why did you change config.guess?  What do you mean by "the
> previous build"?

Because the default for OS X is to include the teeny OS version in the
binary extension dir name. It is about as meaningful as
i686-linux-2.6.19.  Because of that I remove the version number so
that I can rebuild ruby and keep the extensions.

Of course, there are OS changes, even from application point of view.
But I failed to find any correspondence between OS version number
changes and OS interface changes so I just gave up. From my experience
noticeable interface changes tend to correspond to some random teeny
(or patchlevel) version number change while most teeny number changes
bring no changes at all.

Anyway, the interface changes usually apply things like cd burning
software or other stuff that is tightly bound to the kernel and
hardware.

The previous build is the ruby 1.8.5 already installed in the system.
>
> > Apparently the dl extension uses the etc extension in some way,
> > I tried to rename ect to 0etc so that it is built first, and add
> > -I$(extout) to the miniruby command line but it still does not work:
>
> 'etc' is required via fileutils.rb and mkmf.rb.
>
> > ../../miniruby -I'../..' -I'../.././lib' -I../../.ext
> > ../.././ext/dl/mkcallback.rb > callback.func
> > /sw/lib/ruby/1.8/i686-darwin/etc.bundle: [BUG] Bus Error
> > ruby 1.8.6 (2007-03-13) [i686-darwin]
>
> Can't you show the backtrace if core exists?

I will try to find/create some.

Thanks

Michal