New patches: [FIX #1916: don't try to convert float constants to int in CMM optimizer Bertram Felgenhauer **20071122095513] { hunk ./compiler/cmm/CmmOpt.hs 344 - maybe_conversion (MO_S_Conv from _) = Just (from, narrowS) + maybe_conversion (MO_S_Conv from _) + | not (isFloatingRep from) = Just (from, narrowS) } Context: [FIX #1910: fix code generated for GDTOI on x86_32 Bertram Felgenhauer **20071121102627] [Add built-in Double operations to vectorisation Roman Leshchinskiy **20071122002517] [Teach vectorisation about Double Roman Leshchinskiy **20071121054932] [Vectorise polyexprs with notes Roman Leshchinskiy **20071121053102] [Make rebindable do-notation behave as advertised simonpj@microsoft.com**20071121174914 Adopt Trac #1537. The patch ended up a bit bigger than I expected, so I suggest we do not merge this into the 6.8 branch. But there is no funadamental reason why not. With this patch, rebindable do-notation really does type as if you had written the original (>>) and (>>=) operations in desguared form. I ended up refactoring some of the (rather complicated) error-context stuff in TcUnify, by pushing an InstOrigin into tcSubExp and its various calls. That means we could get rid of tcFunResTy, and the SubCtxt type. This should improve error messages slightly in complicated situations, because we have an Origin to hand to instCall (in the (isSigmaTy actual_ty) case of tc_sub1). Thanks to Pepe for the first draft of the patch. ] [Add DEBUG-only flag -dsuppress-uniques to suppress printing of uniques simonpj@microsoft.com**20071116152446 This is intended only for debugging use: it makes it easier to compare two variants without the variations between uniques mattering. (Of course, you can't actually feed the output to the C compiler or assembler and expect anything sensible to happen!) ] [Add -dcore-lint when validating libraries simonpj@microsoft.com**20071105164733] [Fix Trac #1913: check data const for derived types are in scope simonpj@microsoft.com**20071121151428 When deriving an instance, the data constructors should all be in scope. This patch checks the condition. ] [Fix Trac #1909: type of map in docs simonpj@microsoft.com**20071120160152] [Move file locking into the RTS, fixing #629, #1109 Simon Marlow **20071120140859 File locking (of the Haskell 98 variety) was previously done using a static table with linear search, which had two problems: the array had a fixed size and was sometimes too small (#1109), and performance of lockFile/unlockFile was suboptimal due to the linear search. Also the algorithm failed to count readers as required by Haskell 98 (#629). Now it's done using a hash table (provided by the RTS). Furthermore I avoided the extra fstat() for every open file by passing the dev_t and ino_t into lockFile. This and the improvements to the locking algorithm result in a healthy 20% or so performance increase for opening/closing files (see openFile008 test). ] [FIX Trac #1825: standalone deriving Typeable simonpj@microsoft.com**20071120125732 Standalone deriving of typeable now requires you to say instance Typeable1 Maybe which is exactly the shape of instance decl that is generated by a 'deriving( Typeable )' clause on the data type decl. This is a bit horrid, but it's the only consistent way, at least for now. If you say something else, the error messages are helpful. MERGE to 6.8 branch ] [FIX #1715: egregious bug in ifaceDeclSubBndrs simonpj@microsoft.com**20071120111723 ifaceDeclSubBndrs didn't have an IfaceSyn case; but with type families an IfaceSyn can introduce subordinate binders. Result: chaos. The fix is easy though. Merge to 6.8 branch. ] [Always do 'setup makefile' before building each library Simon Marlow **20071120103329 This forces preprocessing to happen, which is necessary if any of the .hsc files have been modified. Without this change, a 'setup makefile' would be required by hand after a .hsc file changed. Fortunately 'setup makefile' isn't much extra work, and I've made it not overwrite GNUmakefile if it hasn't changed, which avoids recalculating the dependencies each time. ] [FIX #1847 (improve :browse! docs, fix unqual) claus.reinke@talk21.com**20071108013147 - add example to docs, explain how to interpret output of `:browse! Data.Maybe` - print unqualified names according to current context, not the context of the target module ] [Track changes to package ndp Roman Leshchinskiy **20071120033716] [Temporary hack for passing PArrays from unvectorised to vectorised code Roman Leshchinskiy **20071120024545] [Bind NDP stuff to [:.:] arrays Roman Leshchinskiy **20071119020302] [Don't treat enumerations specially during vectorisation for the moment Roman Leshchinskiy **20071119013729] [Fix bugs in vectorisation of case expressions Roman Leshchinskiy **20071119013714] [More built-in NDP combinators Roman Leshchinskiy **20071119012205] [New vectorisation built-ins Roman Leshchinskiy **20071118051940] [Fix bug in conversion unvect/vect Roman Leshchinskiy **20071118051926] [Extend built-in vectorisation environments Roman Leshchinskiy **20071118045219] [Fix bug in generation of environments for vectorisation Roman Leshchinskiy **20071118045203] [Add builtin var->var mapping to vectorisation Roman Leshchinskiy **20071118042605] [Extend vectorisation built-in mappings with datacons Roman Leshchinskiy **20071118034351] [Change representation of parallel arrays of enumerations Roman Leshchinskiy **20071118033355] [Add vectorisation-related builtin Roman Leshchinskiy **20071118031513] [Teach vectorisation about Bool Roman Leshchinskiy **20071117042714] [Incomplete support for boxing during vectorisation Roman Leshchinskiy **20071117040739] [Make sure some TyCons always vectorise to themselves Roman Leshchinskiy **20071117040537] [Simple conversion vectorised -> unvectorised Roman Leshchinskiy **20071117023029] [Fix bug in case vectorisation Roman Leshchinskiy **20071117015014] [Vectorisation of algebraic case expressions Roman Leshchinskiy **20071116074814] [More vectorisation-related built-ins Roman Leshchinskiy **20071116061831] [Vectorisation utilities Roman Leshchinskiy **20071116051037] [Add vectorisation built-ins Roman Leshchinskiy **20071116050959] [Fix vectorisation of binders in case expressions Roman Leshchinskiy **20071116021833] [Two small typos in the flags summary (merge to 6.8 branch) simonpj@microsoft.com**20071119134639] [Improve the situation for Trac #959: civilised warning instead of a trace msg simonpj@microsoft.com**20071119122938 This doesn't fix the root cause of the bug, but it makes the report more civilised, and points to further info. ] [FIX Trac #1806: test for correct arity for datacon in infix pattern patch simonpj@microsoft.com**20071119114301 Happily the fix is easy; pls merge ] [Accept x86_64-*-freebsd* as well as amd64-*-freebsd* in configure.ac Ian Lynagh **20071117154502 Patch from Brian P. O'Hanlon ] [Attempt at fixing #1873, #1360 Simon Marlow **20071116152148 I think I figured out a reasonable way to manage the GHCi context, comments welcome. Rule 1: external package modules in the context are persistent. That is, when you say 'import Data.Maybe' it survives over :load, :add, :reload and :cd. Rule 2: :load and :add remove all home-package modules from the context and add the rightmost target, as a *-module if possible. This is as before, and makes sense for :load because we're starting a new program; the old home-package modules don't make sense any more. For :add, it usually does what you want, because the new target will become the context. Rule 3: any modules from the context that fail to load during a :reload are remembered, and re-added to the context at the next successful :reload. Claus' suggestion about adding the "remembered" modules to the prompt prefixed with a ! is implemented but commented out. I couldn't decide whether it was useful or confusing. One difference that people might notice is that after a :reload where there were errors, GHCi would previously dump you in the most recent module that it loaded. Now it dumps you in whatever subset of the current context still makes sense, and in the common case that will probably be {Prelude}. ] [Wibble to fix Trac #1901 (shorten messsage slightly) simonpj@microsoft.com**20071116150341] [Improve links from flag reference to the relevant section; and improve doc of RankN flags simonpj@microsoft.com**20071116145816] [FIX Trac #1901: check no existential context in H98 mode simonpj@microsoft.com**20071116145609] [Improve documentation of data type declarations (Trac #1901) simonpj@microsoft.com**20071116081841] [Change the command-line semantics for query commands Simon Marlow **20071116132046 From the help text: Commands that query the package database (list, latest, describe, field) operate on the list of databases specified by the flags --user, --global, and --package-conf. If none of these flags are given, the default is --global --user. This makes it possible to query just a single database (e.g. the global one without the user one), which needed tricks to accomplish before. ] [use "ghc-pkg latest --global" instead of "ghc-pkg list --simple-output" Simon Marlow **20071116122018 The former now does the right thing: it uses the global database only, and picks the most recent package with the given name. ] [Disallow installing packages whose names differ in case only. Simon Marlow **20071116121153 --force overrides. Requested by Duncan Coutts, with a view to treating package names as case-insensitive in the future. ] [FIX BUILD (with GHC 6.2.x): update .hi-boot file Simon Marlow **20071116101227] [FIX #1828: installing to a patch with spaces in Simon Marlow **20071115155747 We have to pass the path to gcc when calling windres, which itself might have spaces in. Furthermore, we have to pass the path to gcc's tools to gcc. This means getting the quoting right, and after much experimentation and reading of the windres sources I found something that works: passing --use-temp-files to windres makes it use its own implementation of quoting instead of popen(), and this does what we want. Sigh. ] [on Windows, install to a directory with spaces (test for #1828) Simon Marlow **20071115155327] [FIX #1679: crash on returning from a foreign call Simon Marlow **20071115131635 We forgot to save a pointer to the BCO over the foreign call. Doing enough allocation and GC during the call could provoke a crash. ] [Avoid the use of unversioned package dependencies Simon Marlow **20071115103249 Fortunately "ghc-pkg list $pkg --simple-output" is a good way to add the version number. ] [FIX #1596 (remove deprecated --define-name) Simon Marlow **20071114165323 Also remove the old command-line syntax for ghc-pkg, which was not documented. Do not merge. ] [FIX #1837: remove deprecated support for unversioned dependencies (do not merge) Simon Marlow **20071114161044 ] [wibble Pepe Iborra **20071114233356] [Make pprNameLoc more robust in absence of loc information Pepe Iborra **20071114233343] [Try to manage the size of the text rendered for ':show bindings' Pepe Iborra **20071114231601] [Make the Term ppr depth aware Pepe Iborra **20071114183417] [Use paragraph fill sep where possible Pepe Iborra **20071114181233] [Make SpecConstr work again simonpj@microsoft.com**20071115084242 In a typo I'd written env instead of env', and as a result RULES are practically guaranteed not to work in a recursive group. This pretty much kills SpecConstr in its tracks! Well done Kenny Lu for spotting this. The fix is easy. Merge into 6.8 please. ] [Documentation only - fix typo in flags reference Tim Chevalier **20071115055748] [Avoid making Either String an instance of Monad in the Haddock parser David Waern **20071114204050] [FIX 1463 (implement 'ghc-pkg find-module') claus.reinke@talk21.com**20071109162652 - the ticket asks for a module2package lookup in ghc-pkg (this would be useful to have in cabal, as well) - we can now ask which packages expose a module we need, eg, when preparing a cabal file or when getting errors after package reorganisations: $ ./ghc-pkg-inplace find-module Var c:/fptools/ghc/driver/package.conf.inplace: (ghc-6.9.20071106) $ ./ghc-pkg-inplace find-module Data.Sequence c:/fptools/ghc/driver/package.conf.inplace: containers-0.1 - implemented as a minor variation on listPackages (as usual, it would be useful if one could combine multiple queries into one) ] [remove --define-name from the --help usage message (#1596) Simon Marlow **20071114153417 ] [FIX #1837: emit deprecated message for unversioned dependencies Simon Marlow **20071114153010] [Fix #782, #1483, #1649: Unicode GHCi input Simon Marlow **20071114151411 GHCi input is now treated universally as UTF-8, except for the Windows console where we do the correct conversion from the current code page (see System.Win32.stringToUnicode). That leaves non-UTF-8 locales on Unix as unsupported, but (a) we only accept source files in UTF-8 anyway, and (b) UTF-8 is quite ubiquitous as the default locale. ] [Fix build David Waern **20071114125842 I had forgot to update HaddockLex.hi-boot-6, so the build with 6.2.2 failed. This fixes that. ] [FIX Trac 1662: actually check for existentials in proc patterns simonpj@microsoft.com**20071114112930 I'd fixed the bug for code that should be OK, but had forgotten to make the test for code that should be rejected! Test is arrowfail004 ] [FIX Trac 1888; duplicate INLINE pragmas simonpj@microsoft.com**20071114104701 There are actually three things here - INLINE pragmas weren't being pretty-printed properly - They were being classified into too-narrow boxes by eqHsSig - They were being printed in to much detail by hsSigDoc All easy. Test is rnfail048. ] [Run the -frule-check pass more often (when asked) simonpj@microsoft.com**20071114104632] [GHCi debugger: added a new flag, -fno-print-binding-contents Pepe Iborra **20071113174539 The contents of bindings show at breakpoints and by :show bindings is rendered using the same printer that :print uses. But sometimes the output it gives spans over too many lines and the user may want to be able to disable it. ] [Fix Trac 1865: GHCi debugger crashes with :print Pepe Iborra **20071113170113] [Replaced two uses of head b explicit pattern matching Pepe Iborra **20071013113136] [Print binding contents in :show bindings Pepe Iborra **20071006123952] [ Leftovers from the 1st GHCi debugger prototype Pepe Iborra **20071004204718] [Following an indirection doesn't count as a RTTI step Pepe Iborra **20070928091941] [FIX #1653 (partially): add -X flags to completion for :set Simon Marlow **20071113153257] [Merge from Haddock: Add <> for images David Waern **20071112220537 A merge of this patch: Mon Aug 7 16:22:14 CEST 2006 Simon Marlow * Add <> for images Submitted by: Lennart Augustsson Please merge to the 6.8.2 branch. ] [Improve documentation of INLINE, esp its interactions with other transformations simonpj@microsoft.com**20071112160240] [Comment re Trac #1220 simonpj@microsoft.com**20071112154109] [Merge from Haddock: Modify lexing of /../ David Waern **20071112023856 Tue Aug 28 11:19:54 CEST 2007 Simon Marlow * Modify lexing of /../ This makes /../ more like '..', so that a single / on a line doesn't trigger a parse error. This should reduce the causes of accidental parse errors in Haddock comments; apparently stray / characters are a common source of failures. Please merge this to the 6.8.2 branch. ] [Merge from Haddock: allow blank lines inside code blocks David Waern **20071112013439 Tue Jan 9 14:14:34 CET 2007 Simon Marlow * allow blank lines inside a @...@ code block Please merge this to the 6.8.2 branch ] [Merge of a patch from the old Haddock branch: David Waern **20071112013143 Fri Jan 5 12:13:41 CET 2007 Simon Marlow * Fix up a case of extra vertical space after a code block Please merge this to the 6.8.2 branch ] [Remove ex-extralibs from libraries/Makefile Ian Lynagh **20071111213618] [Remove the X11 and HGL libraries from extralibs Ian Lynagh **20071111213447 Don Stewart, X11 maintainer, requested we remove X11, and HGL depends on it on Linux (and we don't try to build HGL on Windows). ] [arrows is no longer an extralib Ian Lynagh **20071027123656] [Turn -fprint-bind-result off by default Ian Lynagh **20071111001126] [TAG 2007-11-11 Ian Lynagh **20071111161540] Patch bundle hash: ca05f6b496bf37995d2450e2ed43a1fcadc6ebef