#7510: Immediate seg-fault on 32-bit windows build
    Reporter:  simonpj           |       Owner:                  
        Type:  bug               |      Status:  new             
    Priority:  highest           |   Milestone:  7.8.1           
   Component:  Compiler          |     Version:  7.6.1           
    Keywords:                    |          Os:  Unknown/Multiple
Architecture:  Unknown/Multiple  |     Failure:  None/Unknown    
  Difficulty:  Unknown           |    Testcase:                  
   Blockedby:                    |    Blocking:                  
     Related:                    |  

Comment(by joeyadams):

 I was able to simplify Simon's standalone testcase quite a bit, and still
 get it to segfault (test-reduced.hs).  The segfault occurs if you build
 without optimization (`-O0`), and occurs on both Windows and Linux 32-bit.
 It does not occur on Linux 64-bit.

 However, there's another symptom: `-ddump-cmm` produces a much larger file
 (roughly 5 MB instead of 250 KB) with GHC HEAD than with 7.4.2 or 7.6.1.
 The size of the `-ddump-cmm` output scales with the number of fields in
 `data DynFlags` (about 35 KB per field).  This symptom occurs on both
 32-bit and 64-bit.

 Speculation: there's a bug in how record update is compiled.  This bug is
 not platform- or architecture-specific, though it manifests itself on x86
 32-bit by clobbering the call stack.

Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/7510#comment:2>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler

Glasgow-haskell-bugs mailing list

Reply via email to