#2437: More accurate package dependencies
---------------------------------+------------------------------------------
    Reporter:  simonmar          |       Owner:             
        Type:  task              |      Status:  new        
    Priority:  normal            |   Milestone:  6.12 branch
   Component:  Compiler          |     Version:  6.8.3      
    Severity:  normal            |    Keywords:             
  Difficulty:  Moderate (1 day)  |    Testcase:             
Architecture:  Unknown           |          Os:  Unknown    
---------------------------------+------------------------------------------
 The problem we want to solve here is that there is currently no
 distinction between compile-time package dependencies and link-time
 package dependencies.  Often a compile-time dependency is also a link-time
 dependency, but there are two instances in which it might not be:

  * Template Haskell: some packages might be required for executing TH code
 at
    compile-time, but the compiled code doesn't require those packages to
 be
    linked in.

  * Attributes: we're considering allowing arbitrary attributes to be
 attached
    to declarations, where the attributes are compile-time Haskell
 expressions.
    The same issue as with TH crops up again here.

 Currently we figure out which packages to link by looking at the
 (transitive closure of the) imports.  It would be better to look at the
 external references of the compiled code; some of the packages referred to
 by imports may not need to be linked.

 Similarly, we should figure out the dependencies of a ``package`` by
 taking the union of the link-time dependencies of its compiled modules.
 This means a small changes to Cabal.

 None of this is particularly hard, and doesn't need any changes to the
 interface file format or package database: we just record fewer package
 dependencies than before.  The only tricky bit is traversing the code to
 figure out what the package dependencies should be.

-- 
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/2437>
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