[GHC] #11958: Improved testing of cross-compiler

[GHC]

#11958: Improved testing of cross-compiler
-------------------------------------+-------------------------------------
        Reporter:  erikd             |                Owner:
            Type:  task              |               Status:  new
        Priority:  low               |            Milestone:  8.2.1
       Component:  Test Suite        |              Version:  7.10.3
      Resolution:                    |             Keywords:  cross-compile
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by erikd):

 @thomie suggested running `make test stage=1`. That failed because for
 some reason `hp2ps` doesn't get build during cross-compilation. If I
 remove the test for that program:

 {{{
 diff --git a/testsuite/mk/boilerplate.mk b/testsuite/mk/boilerplate.mk
 index 077d503..64a7408 100644
 --- a/testsuite/mk/boilerplate.mk
 +++ b/testsuite/mk/boilerplate.mk
 @@ -198,10 +198,10 @@ ifeq "$(shell test -x '$(HSC2HS)' && echo exists)"
 ""
  $(error Cannot find hsc2hs: $(HSC2HS))
  endif

 -$(eval $(call canonicaliseExecutable,HP2PS_ABS))
 -ifeq "$(shell test -x '$(HP2PS_ABS)' && echo exists)" ""
 -$(error Cannot find hp2ps: $(HP2PS_ABS))
 -endif

  $(eval $(call canonicaliseExecutable,HPC))
  ifeq "$(shell test -x '$(HPC)' && echo exists)" ""
 }}}

 then the tests do actually run (x86_64/linux cross compiling to
 armhf/linux):

 {{{
 OVERALL SUMMARY for test run started at Sun Apr 24 20:40:22 2016 AEST
  0:28:29 spent to go through
     5128 total tests, which gave rise to
    12346 test cases, of which
     7884 were skipped

       66 had missing libraries
     4096 expected passes
      157 expected failures

       13 caused framework failures
        9 unexpected passes
      134 unexpected failures
        0 unexpected stat failures
 }}}

 It seems some of these tests failed because the 64 bit `Word` results were
 expected instead of using the 32 bit `Word` expected results.

 This idea does seem promising.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11958#comment:1>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler