GitLab

We'll need to change this if we ever use threads in GHC, but for now it
drops mkSplitUniqSupply from 35% time down below 0.1% when compiling
J.hs from #1136.

Whilst compiling Main.hs with -ddump-stg, ddump-asm and friends
you can how add -ddump-to-file and you'll get the dumps redirected
to Main.dump-stg, Main.dump-asm etc.

Linear register allocator now keeps track of why it spilled certain 
registers. It had to be refactored somewhat to thread the information
through, but the algorithm is unchanged.

Compiling module Foo with -ddrop-asm-stats produces a file
called Foo.dump-asm-stats which will contain increasingly
interesting information.

  
  With this patch, :stepover can effectively appear to step over recursive calls and 
  calls to locally bound functions (in a where clause).
  
  However, when we run out of ticks in the current expression, 
  the illusion vanishes and laziness brings us to the body of the last function 
  we "stepped over". 
  This is not desired at all, it is potentially very confusing.
  As a countermeasure, when this happens :stepover emits a warning

     "Warning: no more breakpoints in this function body, switching to :step"  

:stepover only works lexically locally, in the context of the 
current expression. I have tried to make this point clear
in the users guide with an example. 

We have 4 new flags:

  -fno-gen-manifest
     suppresses creation of foo.exe.manifest

  -fno-embed-manifest
     suppresses embedding of the manifest in the executable

  -pgmwindres
     specify a program to use instead of windres

  -optwindres
     extra options to pass to windres

"windres" is now copied from mingw and included in a binary
distribution.

We now have various flags that need to be passed to gcc, but only if
the local gcc version supports them.  So instead of wiring this
knowledge into ghc when it is built, we now put these "extra gcc
flags" into a text file in $libdir, extra-gcc-flags, which is created
by configure.  It is also created by the configure script of a binary
distribution, so a binary dist is independent of the gcc version used
to build it.

Hopefully I got the binary-dist stuff right, but binary dists aren't
currently working so we'll need to come back and check this.

This appears to cause trouble in at least some shells

Refactored linear allocator into separate liveness annotation and allocation stages.
Added graph coloring allocator, use -fregs-graph to enable.
  New dump flags are
    -ddump-asm-native          -- output of cmm -> native transform.
    -ddump-asm-liveness        -- code annotated with register liveness info
    -ddump-asm-coalesce        -- output of register move coalescing
                                    (this is a separate pass when using the coloring allocator)
                                    (this could change in the future)
    -ddump-asm-regalloc        -- code after register allocation
    -ddump-asm-regalloc-stages -- blocks after each build/spill stage of coloring allocator
    -ddump-asm-conflicts       -- a global register liveness graph in graphviz format 
        
The new register allocator will allocate some registers, but it's not
quite ready for prime-time yet. The spill code generator needs some work...

I'm not sure why this was here, but it makes ghci.o get built twice on
Windows, which can cause the build to fail if it tries to link it
in the middle.

This patch adds a check that each imported module is listed in
the dependencies from the previous interface.  It catches the
following cases and forces recompilation:
   - an exposed package has been upgraded
   - we are compiling with different package flags
   - a home module that was shadowing a package module has been removed
   - a new home module has been added that shadows a package module

I haven't yet added the package timestamping as described in #1372.

  * The correct definition of C-- requires that a procedure not
    'fall off the end'.  The 'never returns' annotation tells us
    if a (foreign) call is not going to return.

    Validated!

- Fixed test "Infix" in the testsuite, which Roman added