GHC panic

[nobody]

This concerns GHC HEAD (6.3) on 20 Jan 2005.

I'm afraid this is not going to be a very useful bug
report,
but I thought I'd file my observations anyway.

When compiling a file AFRPTestsReact.hs, GHC panics:

seshat-257% ghc -c -fglasgow-exts -package lang
-package concurrent -i../src AFRPTestsReact.hs
ghc-6.3: panic! (the `impossible' happened, GHC version
6.3):
        Ix{Int}.index: Index (257) out of range ((0,118))
 
Please report it as a compiler bug to
glasgow-haskell-bugs@haskell.org,
or http://sourceforge.net/projects/ghc/.

I start to work on creating a small test case that will
exhibit
the bug by making a copy of the file under a different
name.

But when I compile that file, with the exact same
command line arguments, just to routinely ensure that I
can recreate the bug before I start modifying things,
everything works fine!

I compile the old file again, and again GHC panics, despite
the files being identical , living in the same
directory, being
compiled with the same command line arguments and
from the very same prompt (same environment variables).

I try a few times, and GHC consistently behaves as
described: old file => panic, any copy of the file =>
compiles fine.

Eventually, I "touch" the old file. And then it
compiles just fine!

Possibly some problem with reading file attributes???

Sorry that I cannot provide more info (e.g.
the exact time and date of the file): I'm no
longer able to recreate the problem.

/Henrik

e-mail: nhn@cs.nott.ac.uk

[simonmar]

Logged In: YES 
user_id=48280

It sounds like you have an old .hi file lying around
generated by a previous (development, unreleased) version of
GHC.  GHC only detects version mismatches with .hi files at
the granularity of release versions (6.2, 6,4 etc.), it
won't detect that a .hi file was created by version
6.3.20050110 and we're at 6.3.20050111, for example. 
Nevertheless, such version mismatches can cause compiler
crashes like the one you describe.

[simonmar]

Logged In: YES 
user_id=48280

Submitter indicates that the explanation below is "possibly
correct" but that it can't be reproduced now in any case. 
Hence, bug closed.