Re: [GHC] #7179: Non-unique variable bindings in the Core

[GHC]

#7179: Non-unique variable bindings in the Core
---+
  Reporter:  stephen70edwards  |  Owner:  
  Type:  bug   | Status:  closed  
  Priority:  normal|  Milestone:  
 Component:  External Core |Version:  7.4.1   
Resolution:  wontfix   |   Keywords:  CoreSyn Unique  
Os:  Unknown/Multiple  |   Architecture:  Unknown/Multiple
   Failure:  Other | Difficulty:  Unknown 
  Testcase:|  Blockedby:  
  Blocking:|Related:  
---+
Changes (by simonpj):

  * status:  new => closed
  * difficulty:  => Unknown
  * resolution:  => wontfix


Comment:

 I'm afraid Core makes no claim to uniqueness.  In general, the Simplifier
 tends to remove shadowing, so it will tend not to produce this:
 {{{
   (\x_3. \x_3. ...x_3...)
 }}}
 But that's only to make life easier for the Simplifier itself; it's not a
 guarantee. And no other pass is required to ensure uniqueness.

 The simplifier does not even try to eliminate uses of the same binder in
 different sub-expressions (your example).

 These would be hard invariants to maintain. If you want to establish such
 an invariant, just before you do something else. you can write a pass that
 does so.

 I recently added this comment to `CoreSyn.lhs`:
 {{{
 Note [Shadowing]
 
 While various passes attempt to rename on-the-fly in a manner that
 avoids "shadowing" (thereby simplifying downstream optimizations),
 neither the simplifier nor any other pass GUARANTEES that shadowing is
 avoided. Thus, all passes SHOULD work fine even in the presence of
 arbitrary shadowing in their inputs.

 In particular, scrutinee variables `x` in expressions of the form
 `Case e x t` are often renamed to variables with a prefix
 "wild_". These "wild" variables may appear in the body of the
 case-expression, and further, may be shadowed within the body.
 }}}
 If you would like to modify it, or add some more clarification, can you
 suggest some words and I'll add them?

 Simon

-- 
Ticket URL: 
GHC 
The Glasgow Haskell Compiler

___
Glasgow-haskell-bugs mailing list
Glasgow-haskell-bugs@haskell.org
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs

[GHC] #7179: Non-unique variable bindings in the Core

[GHC]

#7179: Non-unique variable bindings in the Core
--+-
 Reporter:  stephen70edwards  |  Owner:  
 Type:  bug   | Status:  new 
 Priority:  normal|  Component:  External Core   
  Version:  7.4.1 |   Keywords:  CoreSyn Unique  
   Os:  Unknown/Multiple  |   Architecture:  Unknown/Multiple
  Failure:  Other |   Testcase:  
Blockedby:|   Blocking:  
  Related:|  
--+-
 In the process of using GHC as a library, I found that the new variables
 bound in a core Case may not be unique, which I did not expect (what part
 of "unique" don't you understand?). They're in mutually exclusive branches
 and don't seem to adversely affect the generated code (perhaps another
 unique pass runs).

 It's visible in the external core output, although I discovered the
 problem while using core data structures.

-- 
Ticket URL: 
GHC 
The Glasgow Haskell Compiler

___
Glasgow-haskell-bugs mailing list
Glasgow-haskell-bugs@haskell.org
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs