GitLab

On FreeBSD, one needs use "-z origin" in order to enable resolution
of $ORIGIN in RPATH. Part of #7819.

We now just pass the filename as an argument

We now just pass the output filename as an argument instead

The output location needs to be different for the vanilla and dynamic
ways when doing -dynamic-too. Fixes dynamicToo003.

It doesn't work with --make yet. Trac #7864 filed.

-fspec-inline-join-points was removed in Jan 2010:
99f41975

-fno-hi-version-check was removed in Nov 2003:
98744cef

-dno-black-holing was removed in Aug 2011:
1c2f8953

-fruntime-types was removed in Apr 2008:
8d6bc9bf

-fhardwire-lib-paths was removed in Jan 2008:
1b98179e

This is always "YES".

We had inplace/bin/ghc-cabalghc-cabal

I'm not sure if we want to make this change permanently, but for now it
fixes the unreg build.

I've also removed some redundant special-case code that generated
prototypes for foreign functions. The standard pprTempAndExternDecls
now generates them.

In loopDecls, as far as I can see the globalDecls will always
already be in the environment, so don't add them again.

Previously, stacks were always attributed to CCCS_SYSTEM.  Now, we
attribute them to the CCS when the stack was allocated.  If a stack
grows, new stack chunks inherit the CCS of the old stack.
Signed-off-by: Edward Z. Yang <ezyang@mit.edu>

* We now never report derived-constraint type errors, even
  in the "insolubles".  See Note [Insoluble derived constraints]
  in TcRnTypes.

* The cec_suppress mechanism in TcErrors is refactored a bit so that:
   - We suppress *all* errors in unreachable code (they can be jolly
     confusing)
   - We no longer suppress *all* non-insoluble errors if there are *any
     insolubles anywhere.  Instead we are a bit more refined.
  See Note [Suppressing error messages] in TcErrors

This unwanted unification was happening in the zonking phase
which un-flattens type-function applications (TcMType.zonkFlats,
try_zonk_fun_eq).  The main unifier is careful to make the check,
but I'd forgotten it here.  That in turn led to a very confusing
error message.

Zonking can make implications empty becuase it gets rid
of type-function flattening equalities

Patch from Stephen Blackheath.

Patch from Stephen Blackheath.

timer_create doesn't exist and setitimer doesn't fire on iOS, so we're
using a pthreads-based implementation. It may be to do with
interference with the signals of the debugger. Revisit. See #7723.

I don't think these should be necessary. If something breaks as
a result then we can look at why.