On Fri, Mar 27, 2009 at 09:56:53AM -0400, Alexander Hansen wrote:
> This is the expected behavior.  Something in how our lynx package builds
> relies on behavior from a system utility that differs between the
> built-in version and that from coreutils-default.  (I haven't looked at
> it, but it might be possible to do something different here, such as to
> include full paths so that things like "mv", "cp", and "install" are
> guaranteed to be the system's version).  After lynx is built there is no
> such conflict.
> 
> As the message says, coreutils-default will be removed while lynx is
> built, and then reinstalled afterwards.  If there's a build failure,
> coreutils-default will still be reinstalled.
> 
> Basically you just won't have coreutils-default while you're building
> lynx. 

While building itself, lynx autodetects a bunch of command-line
utilities and hard-codes their paths into lynx itself. If we *allow*
it to see coreutils-default, we have to *force* it to see it on all
systems (so that the build always winds up the same way) and set a
runtime Depends on it so that the package will actually be able to use
those paths it found. Or we could just make sure it *doesn't* see any
coreutils-default commands when it builds, and then we don't force all
users to install that package and keep it installed just to use lynx.
Given that the cu-d commands do behave a bit differently, I really
hate the idea of forcing them into the users' PATH. It's probably
possible to override the lynx path detection (hardcode the apple path
even if cu-d is present)...someone who is bothered by the current
hackish solution is welcome to try to improve it obviously.

dan

-- 
Daniel Macks
[email protected]
http://www.netspace.org/~dmacks


------------------------------------------------------------------------------
_______________________________________________
Fink-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/fink-users

Reply via email to