GitLab

The "Unhelpful" cases are now in a separate type. This allows us to
improve various things, e.g.:
* Most of the panic's in SrcLoc are now gone
* The Lexer now works with RealSrcSpans rather than SrcSpans, i.e. it
  knows that it has real locations and thus can assume that the line
  number etc really exists
* Some of the more suspicious cases are no longer necessary, e.g.
  we no longer need this case in advanceSrcLoc:
      advanceSrcLoc loc _ = loc -- Better than nothing

More improvements can probably be made, e.g. tick locations can
probably use RealSrcSpans too.

I'd naively assumed that the haddock-html field would only use the
$httptopdir variable. Hopefully this will fix the windows build.

It is in the NatM monad, which has DynFlags as part of its state.

For now we panic in the OSUnknown case. It would probably be better
to make the enumerations complete instead, though.

 1) generating Derived FDs as happens for equality superclasses
 2) Kept the optimization of immediately discharging items
    if fundeps cause a match
 3) Restructured top-reactions and interactions with inerts to
    behave similarly to each other.

In particular, (1) fixes ticket #5236.

The second argument to C's memset was passed as a W8 while memset
expects an int.
Signed-off-by: David Terei <davidterei@gmail.com>

We were calling freeArgv with arguments of the wrong type.
But actually, we had our own copy of its loop anyway, so
now we just use that instead.

This means that for example
    ghc -ddump-simpl -ddump-to-file ...
now puts the simpl output in a file, rather htan sending it to stdout.

was [(Module, Maybe ImportDecl)], now it is just [ImportDecl].  So now
":m +A" and "import A" do exactly the same thing in GHCi, and use the
same code paths.

This exposes new constants that can be propagated.

Currently the mini-inliner would only forward substitute assignments
to registers that were used exactly once, to not risk duplicating
computation.  For constants there's no such risk so we always
substitute.  Prior to the change the Cmm

    fn
    {
        bits64 a, b;

        a = 1;
        b = a + a;
        RET_N(b);
    }

would be optimized as

    fn()    { []
            }
        ca: _cb::I64 = 1;
            R1 = _cb::I64 + _cb::I64;
            jump (I64[Sp + 0]) ();
    }

but after it would be optimized as

    fn()    { []
            }
        ca: R1 = 2;
            jump (I64[Sp + 0]) ();
    }

Note that this pass does not deal with the now dead assignment.

the former will fail if libbfd is a static library and has a libz
dependency that we don't know about (#3756)

Fixes more failures on arches without an NCG