GitLab

Update expected output

Comment only

wibble

Add a type-synonym test

Update expected output after recent changes to Read:

  - spaces around '=' in record fields
  - (-a) % b is now printed correctly

MERGE TO STABLE

skip win32001 under Win32 also

make it compile

Add comments explaining why we on purpose force net001 and
net002 to fail under Win32, but don't cover up the failure in
the testscript.

mingw32/win32: tweak back to life

unused

Add deriving record show test

Add test for List.sortBy being a stable sort.

recv now raises EOF; handle it

Add a test for more-than-4-arguments-in-foreign-import-wrapper.  From Sven Panne.

#include "Rts.h" so that this compiles (it shouldn't be required though)

Turns out that we *did* have a test that would have shown up the
DsForeign bug (rev. 1.70 of ghc/compiler/deSugar/DsForeign.lhs) but it
was disabled because it didn't quite work with the current test
framework.

I've hacked the framework a bit and hopefully knocked the test into
shape so we shouldn't get the same problem again.

MERGE TO STABLE after testing in HEAD

Put the source file *before* any extra options for this test on the
command line.  That way we can compile extra files that depend on
things created by compiling the original source (eg. _stub files for
multi-file FFI tests).

update expected output (which was wrong)

Add tcfail109

update expected output

FreeBSD-specific test output (mainly just different default buffer sizes).

Solaris-specific versions of stdout for these tests (the default block
size differs on Solaris, which means that Show-ing a Handle gives
slightly different output).

The Solaris build should have zero unexpected test failures now.

Skip test ffi005 due to non-portability

update expected output (which was wrong)

conc021: update expected output and exit code

Fix 32-bit assumption: Int is not CInt

Finally fix foreign export and foreign import "wrapper" so that
exceptions raised during the call are handled properly rather than
causing the RTS to bomb out.

In particular, calling System.exitWith in a foreign export will cause
the program to terminate cleanly with the desired exit code.  All
other exceptions are printed on stderr (and the program is
terminated).

Details:

GHC.TopHandler.runMain is now called runIO, and has type IO a -> IO a
(previously it had type IO a -> IO (), but that's not general enough
for a foreign export).  The stubs for foreign export and forein import
"wrapper" now automatically wrap the computation in runIO or its dual,
runNonIO.  It turned out to be simpler to do it this way than to do
the wrapping in Haskell land (plain foreign exports don't have
wrappers in Haskell).

Test recursive newtype deriving

Add tc158

Add tc157

GlaExts ==> GHC.Exts

Update expected output

Add test for instance decl context validity

Add one of Alastair Reid's FFI tests, with modifications to get it to
run under GHC.

Turn off conc037 and conc038 for now: these depend on $(GhcThreadedRts).

Sync with new seq story

update to not use the lang package

add another missing end-of-line test for hGetLine

Omit bogus ambiguity report after a no-instance error