containers-0.5.6.0 causes GHC testsuite tests to hang

[hvr]

I just tried updating GHC's container submodule to 0.5.6.0 (b9e4e22) but now a couple of testcases in GHC's testsuite simply hang (and get killed by timeouts)

Here's an excerpt:

=====> sequence001(normal) 4362 of 4363 [0, 10, 1] 
cd ../../libraries/containers/tests-ghc && '/home/hvr/Haskell/GHC/ghc/bindisttest/install   dir/bin/ghc' -fforce-recomp -dcore-lint -dcmm-lint -dno-debug-output -no-user-package-db -rtsopts -fno-warn-tabs -fno-ghci-history -o sequence001 sequence001.hs  -package containers  >sequence001.comp.stderr 2>&1
cd ../../libraries/containers/tests-ghc && ./dataintset001    </dev/null >dataintset001.run.stdout 2>dataintset001.run.stderr
cd ../../libraries/containers/tests-ghc && ./sequence001    </dev/null >sequence001.run.stdout 2>sequence001.run.stderr
Wrong exit code (expected 0 , actual 1 )
Stdout:

Stderr:
sequence001: <<loop>>

*** unexpected failure for sequence001(normal)
Timeout happened...killing process...
Wrong exit code (expected 0 , actual 99 )
Stdout:
11

Stderr:

*** unexpected failure for T3333(normal)
[Errno 21] Is a directory: './ghci/linking/.'
[Errno 22] Invalid argument: './ghci/linking/.'

and here's the summary

Unexpected results from:
TEST="print026 package01e ghci037 sequence001 T706 T3333 ghcilink002 ghcilink003 ghcilink004 ghcilink005 ghcilink006 T8333 ghciprog004 hpc_ghc_ghci T3171 T9872b T9872c T9872a T5837"

OVERALL SUMMARY for test run started at Mon Dec 15 23:59:09 2014 CET
 0:13:22 spent to go through
    4363 total tests, which gave rise to
   16859 test cases, of which
   12834 were skipped

      55 had missing libraries
    3884 expected passes
      67 expected failures

       1 caused framework failures
       0 unexpected passes
      15 unexpected failures
       4 unexpected stat failures

Unexpected failures:
   ../../libraries/containers/tests-ghc  sequence001 [bad exit code] (normal)
   ../../libraries/hpc/tests/ghc_ghci    hpc_ghc_ghci [bad exit code] (normal)
   driver                                T706 [bad exit code] (normal)
   ghci.debugger/scripts                 print026 [bad exit code] (ghci)
   ghci/linking                          T3333 [bad exit code] (normal)
   ghci/linking                          ghcilink002 [bad exit code] (normal)
   ghci/linking                          ghcilink003 [bad exit code] (normal)
   ghci/linking                          ghcilink004 [bad exit code] (normal)
   ghci/linking                          ghcilink005 [bad exit code] (normal)
   ghci/linking                          ghcilink006 [bad exit code] (normal)
   ghci/prog004                          ghciprog004 [bad exit code] (normal)
   ghci/scripts                          ghci037 [bad exit code] (normal)
   ghci/should_run                       T3171 [bad exit code] (normal)
   package                               package01e [stderr mismatch] (normal)
   th                                    T8333 [bad exit code] (normal)

Unexpected stat failures:
   perf/compiler  T5837 [stat not good enough] (normal)
   perf/compiler  T9872a [stat not good enough] (normal)
   perf/compiler  T9872b [stat not good enough] (normal)
   perf/compiler  T9872c [stat not good enough] (normal)

/cc @treeowl and @foxik

[foxik]

I will investigate tomorrow. Also, I will have 0.5.6.1, which in addition to this will fix some warnings I forgot and renames the Typeable.h file, which sometimes aliases with the Typeable.h from base (and generates warnings).

BTW, I am seeing several 'redundant imports' warnings on GHC-head, for example for Applicative(..) from Control.Applicative. I imagine lot of code has to deal with this -- is there a simple way of quieting those warnings while keeping the rest? I could only come up with conditional compilation.

[treeowl]

I'm pretty sure conditional compilation is the only way. The hangs are a nasty surprise, but I don't know if I'll have a chance to investigate them tonight.

[treeowl]

On the same subject, the map/coerce tests I added must only be run when optimization is enabled, and I forgot to even think about that. Also, I don't know how. However, that shouldn't lead to any <<<loop>>> exceptions.

[tibbe]

Why are there tests that only work with optimizations enabled?

As for the imports, we should use conditional compilation.

[foxik]

I have disabled the coercion tests for the time being. They depend on RULES being fired, that it why optimization is needed.

[foxik]

@hvr I am seeing the problem, my stupid fault which manifests only on base 4.8. That is what I get for not using cabal when testing (as then it does not use MIN_VERSION_base). If you want some laugh, look at toList definition in Foldable class for base 4.8 :-)

I will commit today, after I will perform GHC validation.

[tibbe]

Once fixed can you please make a bugfix release?

[foxik]

I was counting with that, now that I have the rights :-)

[hvr]

fyi, I've just added a base < 4.8 constraint to containers-0.5.6.0 to mark that release uninstallable for base-4.8:

http://hackage.haskell.org/package/containers-0.5.6.0/revisions/

[foxik]

Thanks, useful.

[foxik]

@hvr I just released containers-0.5.6.1. The containers testsuite works on ghc 7-{0,4,6,8,head} and I also validated current ghc-master (only package01e.stderr must be updated to accomodate increased containers version).

[treeowl]

Why are there tests that only work with optimizations enabled?

Because they are intended to test whether map/coerce RULES are working
properly.

[hvr]

@foxik thanks; I updated GHC via ghc/ghc@5444560 yesterday (so today's PPA ghc-head build does already carry containers-0.5.6.1)