[GHC] #15071: :set usage in ghci tests breaks platform independence of output

[GHC]

#15071: :set usage in ghci tests breaks platform independence of output
-------------------------------------+-------------------------------------
        Reporter:  bgamari           |                Owner:  RolandSenn
            Type:  bug               |               Status:  patch
        Priority:  high              |            Milestone:  8.8.1
       Component:  Test Suite        |              Version:  8.2.2
      Resolution:                    |             Keywords:  ci-breakage
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:  Windows: make
                                     |  test TESTS="ghci057 T9293"
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):  Phab:5125
       Wiki Page:                    |
-------------------------------------+-------------------------------------
Changes (by RolandSenn):

 * testcase:   => Windows: make test TESTS="ghci057 T9293"
 * status:  new => patch
 * differential:   => Phab:5125


Comment:

 I checked Harbormaster build log 297356
 (https://phabricator.haskell.org/harbormaster/log/view/297356/) of build
 52621 (https://phabricator.haskell.org/harbormaster/build/52621/) from
 Sun, Sep 2, 5:14 PM and found the following:

 - Two ghci tests (''ghci057'' and ''T9293'') fail because of differences
 in the stdout file, as described in this ticket.
   The offending line in stdout is no longer ''-fexternal-dynamic-refs''
 but ''-fghci-leak-check''.
   Both tests specify a ''-fno-ghci-leak-check'' flag, so it's Ok, that
 '':set'' does not list this flag.
   I uploaded a patch (https://phabricator.haskell.org/D5125) that changes
 the expected stdout files.

 - The test ''ghci024'' is working and does no longer fail.

 - Two other ghci tests (''T5975a'' and ''T5975b'') fail while setting up
 the test with ''pre-cmd(touch föøbàr1.hs'').
   I never encounter problems, when I use German Umlauts in file names.
   On my Windows machine both tests run successfully.  I guess this is not
 a GHC error,
   but some issue with your CI machine. Maybe either Windows, MinGW or
 Python is outdated.

   On my machine I'm using:
     - Windows: Edition Windows 10 Home, version 1803, build 17134.228.
     - msys64: 20180531
     - python: 3.6.6

 Most of the many other failing tests are a plugin issues and have nothing
 to do with this ticket!

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/15071#comment:5>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler