#3637: ./configure doesn't understand Gentoo's build/host/target
----------------------------------+-----------------------------------------
  Reporter:  kolmodin             |          Owner:                  
      Type:  bug                  |         Status:  new             
  Priority:  high                 |      Milestone:  6.12.1          
 Component:  Build System         |        Version:  6.12.1 RC1      
Resolution:                       |       Keywords:  regression      
Difficulty:  Unknown              |             Os:  Unknown/Multiple
  Testcase:                       |   Architecture:  Unknown/Multiple
   Failure:  Building GHC failed  |  
----------------------------------+-----------------------------------------
Changes (by asuffield):

 * cc: [email protected] (added)
  * failure:  => Building GHC failed

Comment:

 It's not clear what configure is trying to do with the
 --{host,build,target} arguments here. GNU-style configure scripts are
 supposed to accept GNU-format configuration triples (as used by
 config.{sub,guess}) and direct the configure script to set up for a
 particular mode of cross-compilation. The vendor build scripts pass these
 arguments because their own "cross compile the distribution" automation
 knows more about what's going on. There is a cluster of bugs in
 configure.ac that get this all wrong.

 Option 1 is "reimplement config.sub", which is wrong unless you're trying
 to reinvent autoconf. Options 2 and 3 are essentially the same here, both
 being "we shall not support GNU-style cross compiling", which is probably
 also wrong.

 The cross-compile setup code in configure is currently a mess; the
 autodetection is broken. From eyeballing it, I doubt it works at all. To
 move forward:

  * Are there non-obsolete reasons why configure is this way?
  * Is ghc somehow different from other compilers with regard to cross-
 compilation?
  * Was the 6.10 configure logic anything more than legacy gunk to work
 around bugs in autoconf versions from 10 years ago, which are no longer
 relevant?
  * Is there any reason why the whole lot shouldn't be thrown out in favour
 of the normal autoconf handling?

 If the answer to all of those is 'no', the solution is trivial. Otherwise,
 the answers should prove illuminating in how to proceed.

-- 
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/3637#comment:4>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler
_______________________________________________
Glasgow-haskell-bugs mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs

Reply via email to