GHC import bug (# 231631)

Simon Marlow simonmar at microsoft.com
Mon Oct 22 06:07:32 EDT 2001


> I'm sending this to the list, because I did not get much 
> satisfaction from the
> SourceForge bug-tracking tool.
> 
> I have just encountered a GHC bug similar to the one numbered 
> 231631 on
> SourceForge.
> 
> I have some further observations about it, and a small tar 
> file (attached)
> showing how to reproduce a stronger form of it.
> 
> The bug is still present in 5.02, and it does not have to 
> involve the current
> directory - any directory in the import path will do. Nor do I see any
> connection to the re-compilation system, or package manager.

Ok, I propose to change the story to this:

   In single-file-compilation mode (ie. ghc -c), we search
   for .hi files *only* when locating imported modules.

this will have the slightly unfortunate consequence that when compiling
a module in GHCi you might end up importing different modules than when
the module is compiled using ghc -c, but it is a more sensible thing to
do in ghc -c mode.

Cheers,
	Simon




More information about the Glasgow-haskell-bugs mailing list