GitLab

companion to Lexer.x rev. 1.29; columns now always count tabs as 1 in SrcLoc.

- include the contents of ghc/HACKING verbatim, as a "quick start"
  guide

- change the title to "Building and devleloping GHC", and make it a
  bit more GHC-centric.

add info about actually doing the build

Add a "getting started" guide to hacking on GHC.

Mac OS X/Darwin PowerPC: Fix a problem introduced by the recent Darwin/x86
commit.

Remember how we deliberately misaligned the .o file in memory to compensate
for Mach-O's lax alignment rules.
This should have been comitted along with Linker.c 1.203 about two weeks ago.

define THREADED_RTS for the SMP way too

Fix the build; none of this is actually used in THREADED_RTS or SMP,
but I haven't got around to fixing things so we can omit it from the
build completely yet.

Fix a warning

Fix build for way "u"

more Win32 fixes

Fix Win32 DEBUG warnings

Fix Win32 build

No need to import platform-specific stuff now.

Itimer.h has moved into posix/

wibble

oops, remove old file

wibble

Big re-hash of the threaded/SMP runtime

This is a significant reworking of the threaded and SMP parts of
the runtime.  There are two overall goals here:

  - To push down the scheduler lock, reducing contention and allowing
    more parts of the system to run without locks.  In particular,
    the scheduler does not require a lock any more in the common case.

  - To improve affinity, so that running Haskell threads stick to the
    same OS threads as much as possible.

At this point we have the basic structure working, but there are some
pieces missing.  I believe it's reasonably stable - the important
parts of the testsuite pass in all the (normal,threaded,SMP) ways.

In more detail:

  - Each capability now has a run queue, instead of one global run
    queue.  The Capability and Task APIs have been completely
    rewritten; see Capability.h and Task.h for the details.

  - Each capability has its own pool of worker Tasks.  Hence, Haskell
    threads on a Capability's run queue will run on the same worker
    Task(s).  As long as the OS is doing something reasonable, this
    should mean they usually stick to the same CPU.  Another way to
    look at this is that we're assuming each Capability is associated
    with a fixed CPU.

  - What used to be StgMainThread is now part of the Task structure.
    Every OS thread in the runtime has an associated Task, and it
    can ask for its current Task at any time with myTask().

  - removed RTS_SUPPORTS_THREADS symbol, use THREADED_RTS instead
    (it is now defined for SMP too).

  - The RtsAPI has had to change; we must explicitly pass a Capability
    around now.  The previous interface assumed some global state.
    SchedAPI has also changed a lot.

  - The OSThreads API now supports thread-local storage, used to
    implement myTask(), although it could be done more efficiently
    using gcc's __thread extension when available.

  - I've moved some POSIX-specific stuff into the posix subdirectory,
    moving in the direction of separating out platform-specific
    implementations.

  - lots of lock-debugging and assertions in the runtime.  In particular,
    when DEBUG is on, we catch multiple ACQUIRE_LOCK()s, and there is
    also an ASSERT_LOCK_HELD() call.

What's missing so far:

  - I have almost certainly broken the Win32 build, will fix soon.

  - any kind of thread migration or load balancing.  This is high up
    the agenda, though.

  - various performance tweaks to do

  - throwTo and forkProcess still do not work in SMP mode

Column numbers in SrcLocs are now counted as the number of characters,
rather than columns.  i.e. a tab always counts as 1.  This was
necessary for communication with Visual Studio interfaces which expect
character indices, but also it seems the majority of other compilers
also do things this way.

From: Krasimir Angelov <kr.angelov@gmail.com>

changes to exitScheduler(): instead of waiting for all the tasks to
stop, which is unreasonable, we just wait for the run queue to drain.
This is much quicker, but not ideal (see comments).

[mingw]delay flushing console buffer until last possible moment. Merge to STABLE

ocGetNames_PEi386(): correctly allocate .bss sections in relocatable
 object files. See commit for details; fixes misc GHCi crashes reported
 against 6.4.1.

Merge to STABLE.

For the 100th time: Fixed the DocBook XML. >:-( Please, please, please:
Everybody should do a "make validate" before checking in any XML stuff!

Buglets in GADT record-syntax stuff, which killed the weekend builds

Small simplifier bug in case optimisation
	~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

The simplifier eliminates redundant case branches, and panics if there
are no case alternatives.  But due to a slightly delayed instantiation
of a type constructor variable 'p' by a type constructor 'P', it turned
out that an inner case had no alternatives at all, becuase an outer case
had not pruned a branch as quickly as it should have.

This commit fixes both problems:

a) SimplUtils.mkCase1 now returns a call to 'error' (instead of panicing)
   when it gets an empty list of alternatives.   Somewhat analogous to
   the inaccessible GADT case in Simplify.simplifyAlt

b) In SimplUtils.prepareDefault, use the up-to-date scrutinee, rather than
   the less up-to-date case_bndr, to get the case type constructor.  That
   leads to slightly earlier pruning of inaccessible branches.

Fixes a bug reported by Ian Lynagh.

Test is simplCore/should_compile/simpl013

Comments only

merge rev. 1.13.2.12 from ghc-6-4-branch

Fix a comment

note about watching for/preventing overwrites of critical files in the
bootstrap process.

if TARGETPLATFORM differs from HOSTPLATFORM, don't attempt to build
DerivedConstants,h, ghcautoconf.h and GHCConstants.h.  If these aren't
present, emit a message to remind the user to copy them from the
target system.  Hopefully this should make bootstrapping slightly less
error prone.

StrHash doesn't appear to be used; remove it.  I think it was an
earlier version of the string hashing code in Hash.c.

Some more informative diagnostics for ghc -v about what the package
system is doing.  This should help when diagnosing strange-looking
errors from GHC:

Using package config file: /home/simonmar/fp/lib/i386-unknown-linux/ghc-6.4.1/package.conf
package posix-1.0 will be ignored due to missing dependencies:
  lang-1.0
package util-1.0 will be ignored due to missing dependencies:
  lang-1.0
package data-1.0 will be ignored due to missing dependencies:
  lang-1.0
package text-1.0 will be ignored due to missing dependencies:
  lang-1.0
package Cabal-1.1.4 will be ignored due to missing dependencies:
  util-1.0
*** Deleting temp files
Deleting:
ghc-6.4.1: unknown package: Cabal-1.1.4

Fix a couple of problems with the "unknown package" error message

Add record syntax for GADTs
	~~~~~~~~~~~~~~~~~~~~~~~~~~~

Atrijus Tang wanted to add record syntax for GADTs and existential
types, so he and I worked on it a bit at ICFP.  This commit is the
result.  Now you can say

 data T a where
  T1 { x :: a }           	 :: T [a]
  T2 { x :: a, y :: Int } 	 :: T [a]
  forall b. Show b =>
 	T3 { naughty :: b, ok :: Int } :: T Int
  T4 :: Eq a => a -> b -> T (a,b)

Here the constructors are declared using record syntax.

Still to come after this commit:
  - User manual documentation
  - More regression tests
  - Some missing cases in the parser (e.g. T3 won't parse)
Autrijus is going to do these.


Here's a quick summary of the rules.  (Atrijus is going to write
proper documentation shortly.)

Defnition: a 'vanilla' constructor has a type of the form
	forall a1..an. t1 -> ... -> tm -> T a1 ... an
No existentials, no context, nothing.  A constructor declared with
Haskell-98 syntax is vanilla by construction.  A constructor declared
with GADT-style syntax is vanilla iff its type looks like the above.
(In the latter case, the order of the type variables does not matter.)

* You can mix record syntax and non-record syntax in a single decl

* All constructors that share a common field 'x' must have the
  same result type (T [a] in the example).

* You can use field names without restriction in record construction
  and record pattern matching.

* Record *update* only works for data types that only have 'vanilla'
  constructors.

* Consider the field 'naughty', which uses a type variable that does
  not appear in the result type ('b' in the example).  You can use the
  field 'naughty' in pattern matching and construction, but NO
  SELECTOR function is generated for 'naughty'.  [An attempt to use
  'naughty' as a selector function will elicit a helpful error
  message.]

* Data types declared in GADT syntax cannot have a context. So this
is illegal:
	data (Monad m) => T a where
		  ....

* Constructors in GADT syntax can have a context (t.g. T3, T4 above)
  and that context is stored in the constructor and made available
  when the constructor is pattern-matched on.  WARNING: not competely
  implemented yet, but that's the plan.



Implementation notes
~~~~~~~~~~~~~~~~~~~~
- Data constructors (even vanilla ones) no longer share the type
  variables of their parent type constructor.

- HsDecls.ConDecl has changed quite a bit

- TyCons don't record the field labels and type any more (doesn't
  make sense for existential fields)

- GlobalIdDetails records which selectors are 'naughty', and hence
  don't have real code.

add protos for HeapStackCheck.cmm:stg_block_blackhole_* entry points

- we should be acquiring the sm_lock around allocGroup()
- sanity check threads as they are pulled off the blackhole_queue

MERGE TO STABLE

Fix a bug in TcUnify.unifyTyConApp that made a GADT program fail.
The trouble happens if the type that we are expecting to be
a TyConApp is of form (m a b), where 'm' is refined to a type
constructor.  Then we want to get nice rigid results, and we weren't.

Small refactoring

DEBUG_FILL_SLOP: don't do anything on SMP, zeroing slop words isn't safe

checkHeap: just return on SMP, we can't do heap sanity checking
because we can't zero slop words safely.