"ld: atom sorting error for ..." on OS X

[igloo]

I'm seeing atom sorting errors while building GHC on OS X, e.g.:

Configuring ghc-bin-6.9...
/Users/ian/ghc/darcs/val/libraries/cabal-bin /usr/bin/ghc /Users/ian/ghc/darcs/val/libraries/bootstrapping.conf build     --distpref dist-stage2 
Preprocessing executables for ghc-bin-6.9...
Building ghc-bin-6.9...
[1 of 1] Compiling Main             ( Main.hs, dist-stage2/build/ghc/ghc-tmp/Main.o )
Linking dist-stage2/build/ghc/ghc ...
ld: atom sorting error for _ghczm6zi9_LibFFI_Czuffizutype_closure_tbl and _ghczm6zi9_LibFFI_Czuffizucif_closure_tbl in /Users/ian/ghc/darcs/val/compiler/dist-stage2/build/libHSghc-6.9.a(LibFFI.o)
ld: atom sorting error for _ghczm6zi9_LibFFI_Czuffizutype_closure_tbl and _ghczm6zi9_LibFFI_Czuffizucif_closure_tbl in /Users/ian/ghc/darcs/val/compiler/dist-stage2/build/libHSghc-6.9.a(LibFFI.o)

It doesn't seem to cause any problems, but it's a little worrying.

[judah]

The following post gives a little information about this issue:

 http://www.haskell.org/pipermail/glasgow-haskell-users/2008-January/013990.html

General consensus is that it's harmless; but it would be nice to eliminate the messages if possible.

[chak]

It may be harmless, but it is certainly annoying as it makes tests in the testsuite fail for reasons unrelated to the test (with the usual implications for validate).

[guest]

You can avoid the warnings by passing the -w flag to the linker, i.e., -optl -w.

[igloo]

This is actually a pretty serious problem in practice, because end users assume that compilation failed when they see it.

[colin-adams]

Seconded. I didn't attempt to run my program last night (on a laptop - offline from the internet) - I assumed I would have to report the bug and await a fix. Fortunately, I googled first and found this bug.

[chak]

FWIW, I don't think that I have seen this warning on Snow Leopard at all.

[igloo]

Testcase:

module Main (main) where

data MyType
data MyType2

main :: IO ()
main = print ()

$ ghc-stage2 --make q -XEmptyDataDecls -fforce-recomp
[1 of 1] Compiling Main             ( q.hs, q.o )
Linking q ...
ld: atom sorting error for _Main_MyType_closure_tbl and _Main_MyType2_closure_tbl in q.o
ld: atom sorting error for _Main_MyType_closure_tbl and _Main_MyType2_closure_tbl in q.o
ld: atom sorting error for _Main_MyType_closure_tbl and _Main_MyType2_closure_tbl in q.o

[igloo]

Fixed by

Sat Feb 27 17:39:51 GMT 2010  Ian Lynagh <igloo@earth.li>
  * Fix trac #2578
  We define empty datatypes as not being enumerations, which means the
  empty blocks aren't generated.

in HEAD and 6.12.