Skip to content
Snippets Groups Projects
Commit 345d7805 authored by Ian Lynagh's avatar Ian Lynagh
Browse files

More on release notes

parent bbf8aa55
No related branches found
No related tags found
No related merge requests found
Hide whitespace changes
Inline Side-by-side
docs/users_guide/7.4.1-notes.xml
+ 1045
− 6
View file @ 345d7805
......@@ -9,14 +9,1054 @@
</para>
<sect2>
<title>TODO</title>
<title>Highlights</title>
<para>
The highlights, since the 7.0 branch, are:
</para>
<itemizedlist>
<listitem>
<para>
TODO
There is a new feature Safe Haskell
(<literal>-XSafe</literal>,
<literal>-XTrustworthy</literal>,
<literal>-XUnsafe</literal>):
<xref linkend="safe-haskell" />.
The design has changed since 7.2.
</para>
</listitem>
<listitem>
<para>
There is a new feature kind polymorphism
(<literal>-XPolyKinds</literal>):
<xref linkend="kind-polymorphism-and-promotion" />.
A side-effect of this is that, when the extension is not
enabled, in certain circumstances kinds are now defaulted to
<literal>*</literal>
rather than being inferred.
</para>
</listitem>
<listitem>
<para>
There is a new feature constraint kinds (-XConstraintKinds):
<xref linkend="constraint-kind" />.
</para>
</listitem>
<listitem>
<para>
It is now possible to give any sort of declaration at the
ghci prompt:
<xref linkend="ghci-decls" />.
</para>
</listitem>
<listitem>
<para>
The profiling and hpc implementations have been merged and
overhauled. Visible changes include renaming of profiling
flags and the cost-centre stacks have a new semantics, which
should in most cases result in more useful and intuitive
profiles. The <literal>+RTS -xc</literal> flag now also gives
a stack trace.
</para>
</listitem>
<listitem>
<para>
It is now possible to write compiler plugins:
<xref linkend="compiler-plugins" />.
</para>
</listitem>
<listitem>
<para>
DPH support has been significantly improved.
</para>
</listitem>
<listitem>
<para>
There is now preliminary support for registerised
compilation on the ARM platform, using LLVM.
</para>
</listitem>
</itemizedlist>
<sect2>
<title>Full details</title>
<sect3>
<title>Language</title>
<itemizedlist>
<listitem>
<para>
GHC previously accepted this code:
</para>
<programlisting>
data T f = T
class C a where
t :: f a -> T f
</programlisting>
<para>
inferring the kind of <literal>f</literal> as
<literal>* -> *</literal>, but it now (correctly,
according to the Haskell 98 and Haskell 2010
standards) rejects it. You need to write this
instead:
</para>
<programlisting>
{-# LANGUAGE KindSignatures #-}
data T (f :: * -> *) = T
class C a where
t :: f a -> T f
</programlisting>
</listitem>
<listitem>
<para>
There is a new extension
<literal>PolyKinds</literal>.
This provides 2 related new features:
</para>
<para>
Firstly, suitable user-defined datatypes are now
automatically "promoted" to kinds, e.g.
<literal>Nat</literal> here:
</para>
<programlisting>
data Nat = Zero | Succ Nat
data Vector :: * -> Nat -> * where
VNil :: Vector a Zero
VCons :: a -> Vector a n -> Vector a (Succ n)
</programlisting>
<para>
And secondly, it is now possible for kinds to be
polymorphic. For example, instead of
</para>
<programlisting>
class Typeable (t :: *) where
typeOf :: t -> TypeRep
class Typeable1 (t :: * -> *) where
typeOf1 :: t a -> TypeRep
[...]
instance Typeable Int where typeOf _ = TypeRep
instance Typeable1 [] where typeOf _ = TypeRep
</programlisting>
<para>
you can now say
</para>
<programlisting>
data Proxy t = Proxy
class Typeable t where
typeOf :: Proxy t -> TypeRep
instance Typeable Int where typeOf _ = TypeRep
instance Typeable [] where typeOf _ = TypeRep
</programlisting>
<para>
and the kind of <literal>Proxy</literal> is
polymorphic:
<literal>forall k. k -> *</literal>.
</para>
<para>
See
<xref linkend="kind-polymorphism-and-promotion" />
for more information.
</para>
</listitem>
<listitem>
<para>
The Safe Haskell feature, new in GHC 7.2, has been
redesigned in GHC 7.4. The motivation was to stop
Safe Haskell from causing compilation failures for
people not interested in using it.
</para>
<para>
GHC now tries to infer whether a module is
safe, unless the new
<literal>-fno-safe-infer</literal> flag is given.
Therefore, as well as the old
<literal>-XSafe</literal>, there is now a
<literal>-XUnsafe</literal>
flag to explicitly state that a module should be
treated as unsafe. The old
<literal>-XSafeImports</literal> has been removed.
</para>
<para>
The new flags <literal>-fwarn-safe</literal> and
<literal>-fwarn-unsafe</literal> give warnings
when a module is inferred to be safe or unsafe,
respectively.
</para>
<para>
There is a new flag
<literal>-fpackage-trust</literal>. This controls
whether packages containing imported trustworthy
modules must be marked as
<literal>trusted</literal>.
</para>
</listitem>
<listitem>
<para>
There is now a <literal>NOUNPACK</literal>
pragma, which does the opposite of the existing
<literal>PACK</literal> pragma. It is mainly useful
when <literal>-funbox-strict-fields</literal> has
been used, allowing you to declare that certain
fields should not be unpacked.
</para>
</listitem>
<listitem>
<para>
GHC now requires, as per the standard, that
if a <literal>newtype</literal> is used in an
FFI declaration, then the constructor for that
type must be in scope. For now you only get a
warning if it is not, but in the future this will
be an error.
</para>
</listitem>
<listitem>
<para>
There is a new extension
<literal>ConstraintKind</literal>
which adds a new kind, called
<literal>Constraint</literal>,
to GHC's type system. Then, for example,
</para>
<programlisting>
Show :: * -> Constraint
(?x::Int) :: Constraint
(Int ~ a) :: Constraint
</programlisting>
<para>
You can now write <em>any</em> type with kind
<literal>Constraint</literal> on the left of
<literal>=&gt;</literal>, i.e. you can use type
synonyms, type variables, indexed types, etc.
</para>
</listitem>
<listitem>
<para>
It is now possible to derive an
<literal>Eq</literal> instance for types with no
constructors.
</para>
</listitem>
<listitem>
<para>
In the <literal>MonadComprehensions</literal>
extension, the
<literal>then group by e</literal>
form has been removed. You now need to explicitly
say
<literal>then group by e using groupWith</literal>.
</para>
</listitem>
<listitem>
<para>
There is a new extension
<literal>TraditionalRecordSyntax</literal> which is
on by default. When turned off, the standard Haskell
record syntax cannot be used.
</para>
</listitem>
<listitem>
<para>
In DPH, it is now possible to vectorise things
imported from other modules.
</para>
</listitem>
<listitem>
<para>
In DPH, the <literal>VECTORISE</literal> and
<literal>VECTORISE SCALAR</literal> pragmas now have
<literal>type</literal>, <literal>class</literal>
and <literal>instance</literal> variants.
See
<ulink url="http://hackage.haskell.org/trac/ghc/wiki/DataParallel/VectPragma">VectPragma</ulink> for more details.
</para>
</listitem>
<listitem>
<para>
The <literal>-fdph-seq</literal>,
<literal>-fdph-par</literal>,
<literal>-fdph-this</literal> and
<literal>-fdph-none</literal> flags have been
removed. The vectoriser is now controlled by which
<literal>Data.Array.Parallel</literal> and
<literal>Data.Array.Parallel.Prim</literal>
modules are in scope.
</para>
</listitem>
<listitem>
<para>
GHC now warns consistently about unused type
variables.
</para>
</listitem>
<listitem>
<para>
GHC now implements the static pattern semantics
as clarified by the Haskell' committee, i.e.
the binding <literal>p = e</literal> is now
equivalent to
</para>
<programlisting>
t = e
f = case t of p -> f
g = case t of p -> g
</programlisting>
<para>
where <literal>f</literal> and <literal>g</literal>
are the variables bound by <literal>p</literal>.
</para>
</listitem>
<listitem>
<para>
The <literal>MonoPatBinds</literal>
extension is now deprecated, and has no effect.
</para>
</listitem>
<listitem>
<para>
GHC will now reject a declaration if it infers a
type for it that is impossible to use unambiguously.
For example, given
</para>
<programlisting>
class Wob a b where
to :: a -> b
from :: b -> a
foo x = [x, to (from x)]
</programlisting>
<para>
GHC would infer the ambiguous type
</para>
<programlisting>
foo :: forall a b. Wob a b => b -> [b]
</programlisting>
<para>
but it is impossible to use
<literal>foo</literal> as
<literal>a</literal> will always be ambiguous,
so the declaration is rejected.
</para>
</listitem>
<listitem>
<para>
It is now possible for associated types to have
fresh parameters, e.g. this is now allowed:
</para>
<programlisting>
class C a where
type T a b :: *
instance C Int
type T Int b = b -> b
</programlisting>
<para>
where <literal>T</literal> has a type index
<literal>b</literal> that is not one of the class
variables.
</para>
</listitem>
<listitem>
<para>
It is now possible for multiple associated type
declarations to be given in a single instance,
e.g.
</para>
<programlisting>
class C a where
type T a x :: *
data A
data B
instance C Int where
type T Int A = Int
type T Int B = Bool
</programlisting>
</listitem>
<listitem>
<para>
The import and export of type family data
constructors has been refined. You now need to be
more explicit about what should be exported.
</para>
</listitem>
<listitem>
<para>
Associated type default declarations are now
fully supported. These allow you to specify
a default definition for a type that will be used
if an instance doesn't define its own type, e.g.
</para>
<programlisting>
class Cls a where
type Typ a
type Typ a = Maybe a
instance Cls Int where
</programlisting>
<para>
See <xref linkend="assoc-decl-defs" /> for more
information.
</para>
</listitem>
<listitem>
<para>
You can now specify what simplifier phases
<literal>SPECIALISE</literal> pragmas should be
applied in, in the same way that you could for
<literal>RULE</literal> pragmas, e.g. to specialise
<literal>foo</literal> only in phase 1 use:
</para>
<programlisting>
{-# SPECIALISE [1] foo :: Int -> Int #-}
</programlisting>
</listitem>
<listitem>
<para>
The typechecker is now able to do full constraint
solving under a for-all, i.e. it can cope with the
unification
<literal>forall a. t1 ~ forall a. t2</literal>.
</para>
</listitem>
<listitem>
<para>
The kind <literal>?</literal> has been renamed to
<literal>OpenKind</literal>, and
<literal>??</literal> to <literal>ArgKind</literal>.
</para>
</listitem>
</itemizedlist>
</sect3>
<sect3>
<title>Compiler</title>
<itemizedlist>
<listitem>
<para>
The recompilation checker now takes into account
what flags were used when compiling. For example,
if you first run <literal>ghc -c Foo.hs</literal>,
and then <literal>ghc -DBAR -c Foo.hs</literal>,
then GHC will now recompile
<literal>Foo.hs</literal>.
</para>
</listitem>
<listitem>
<para>
The recompilation checker now also tracks files that
are <literal>#include</literal>d in Haskell sources.
</para>
<para>
Note that we still don't get
<literal>#include</literal>d files in the
<literal>ghc -M</literal> output.
</para>
</listitem>
<listitem>
<para>
The simplifier now maintains a count of how much
transformation it does, and prints a warning if
that exceeds a limit defined by the new
<literal>-fsimpl-tick-factor=N</literal> flag
(default is 100). The intention is to detect when
it seems likely that GHC has gone into an infinite
loop. In the future, GHC may give an error when the
limit is exceeded.
</para>
</listitem>
<listitem>
<para>
There is a new flag
<literal>-fpedantic-bottoms</literal> which makes
GHC be more precise about its treatment of bottoms.
In particular, it stops GHC eta-expanding through
case expressions (which means performance can be
worse).
</para>
</listitem>
<listitem>
<para>
GHC now knows how to call gcc to compile
objective-c++ (<literal>.mm</literal> or
<literal>.M</literal> files).
</para>
</listitem>
<listitem>
<para>
The <literal>-optm</literal>
flag, which allowed extra arguments to be passed to
the mangler, is now deprecated and does nothing.
It will be removed in a future release.
</para>
</listitem>
<listitem>
<para>
GHC now works with LLVM version 3.0, and requires at
least version 2.8.
</para>
</listitem>
<listitem>
<para>
We now pass gcc
<literal>-no_compact_unwind</literal>
on OS X x86 (as well as x86_64), in order to avoid
the
</para>
<programlisting>
ld: warning: could not create compact unwind for .LFB3: non-standard register 5 being saved in prolog
</programlisting>
<para>
warning.
</para>
</listitem>
<listitem>
<para>
The context stack depth, which defines how deeply
the type constraint solver may reason, has been
increased from 20 to 200, as some people were
running into the limit.
</para>
</listitem>
<listitem>
<para>
On x86, the new <literal>-msse4.2</literal>
flag tells GHC that it may use SSE4.2 instructions.
</para>
</listitem>
<listitem>
<para>
There is a new flag
<literal>-dno-llvm-mangler</literal>
which means the LLVM mangler isn't run. It is mainly
useful when debugging GHC.
</para>
</listitem>
<listitem>
<para>
There is a new flag
<literal>-dsuppress-var-kinds</literal>
which can make the output clearer when
<literal>-dppr-debug</literal> is also used.
</para>
</listitem>
<listitem>
<para>
The <literal>-keep-llvm-files</literal>
flag now implies <literal>-fllvm</literal>.
</para>
</listitem>
<listitem>
<para>
The <literal>-split-objs</literal> flag can now be
used with the LLVM backend.
</para>
</listitem>
<listitem>
<para>
There is a new flag <literal>-dumpdir</literal>
which allows you to specify the directory in which
the output of the <literal>-ddump-*</literal> flags
should be put when <literal>-ddump-to-file</literal>
is used. The <literal>-outputdir</literal> flag will
now also set the dump directory.
</para>
</listitem>
<listitem>
<para>
Bitrotted registerised ports for mips, ia64, alpha,
hppa1 and m68k have been removed.
</para>
</listitem>
</itemizedlist>
</sect3>
<sect3>
<title>GHCi</title>
<itemizedlist>
<listitem>
<para>
It is now possible to give any top-level
declaration at the GHCi prompt, e.g.
</para>
<programlisting>
Prelude> data D = C Int
Prelude> let f (C i) = i in f (C 5)
5
</programlisting>
<para>
The current set of declarations are shown by
the new
<literal>:show bindings</literal>
command. See
<xref linkend="ghci-decls" />.
for more information
</para>
</listitem>
<listitem>
<para>
There is a new GHCi command
<literal>:kind!</literal>
which is like
<literal>:kind</literal>
except it also prints the normalised type; e.g., given
</para>
<programlisting>
type family F a
type instance F Int = Bool
</programlisting>
<para>
we get
</para>
<programlisting>
*Main> :kind F Int
F Int :: *
*Main> :kind! F Int
F Int :: *
= Bool
</programlisting>
</listitem>
<listitem>
<para>
There is a new flag
<literal>-fno-ghci-history</literal>
which stops GHCi from loading and saving the GHCi
command history from and to
<literal>~/.ghc/ghci_history</literal>.
</para>
</listitem>
<listitem>
<para>
Library loading in GHCi has been improved; in
particular, loading
<literal>libstdc++.so</literal>
now works, and GHCi will use
<literal>.a</literal> archives to satisfy
<literal>-lfoo</literal> flags on its commandline.
</para>
</listitem>
<listitem>
<para>
When using
<literal>:load</literal>, GHCi will not unload
the current modules until it has successfully loaded
the new ones. This fixes this old problem:
</para>
<programlisting>
Prelude> :l foo
target `foo' is not a module name or a source file
>
</programlisting>
</listitem>
</itemizedlist>
</sect3>
<sect3>
<title>Template Haskell</title>
<itemizedlist>
<listitem>
<para>
GHC used to treat <literal>\|</literal>
and <literal>\]</literal> within quasiquote as
escaped characters, parsing them as
<literal>|</literal> and <literal>]</literal>
respectively. It now does not treat anything
specially; if you would like to be able to include
sequences such as <literal>|]</literal> inside a
quasi-quote then you must define your own escaping
mechanism in the quasi-quoter.
</para>
</listitem>
<listitem>
<para>
The interaction between the recompilation checker
and Template Haskell has been improved. We now
track which modules are used by splices, and if
a module is changed then any splices that use it
will need to be rerun.
</para>
</listitem>
</itemizedlist>
</sect3>
<sect3>
<title>Profiling</title>
<itemizedlist>
<listitem>
<para>
The profiling, coverage (HPC) and breakpoints
infrastructure has been overhauled, and the
three systems are now much more unified.
</para>
<para>
As a result of these changes, the cost-centre
stacks have a new semantics, which should give
more intuitive profiling results. HPC is also
slightly improved: In particular, unused derived
instances and record field names are now
highlighted, and entry counts are more accurate.
</para>
<para>
The <literal>+RTS -xc</literal> flag now also gives
a stack trace.
</para>
<para>
The
<literal>-auto-all</literal> flag has been renamed
to <literal>-fprof-auto</literal>,
<literal>-auto</literal> renamed to
<literal>-fprof-auto-exported</literal>, and
<literal>-caf-all</literal> renamed to
<literal>-fprof-cafs</literal>. The old names are
also still accepted for now.
</para>
<para>
There are also two new flags.
The <literal>-fprof-auto-top</literal> flag
annotates all top-level bindings with SCCs, and
the <literal>-fprof-auto-calls</literal> flag
adds SCCs to all applications. This last flag
is particularly useful for stack traces.
</para>
<para>
Another new flag
<literal>-fprof-no-count-entries</literal>
indicates that entry counts do not need to be
maintained. This makes profiled code faster, and
is particularly useful when heap profiling, as
heap profiles do not use the entry counts.
</para>
</listitem>
<listitem>
<para>
Cost centre names are now in UTF8 rather than
Latin-1.
</para>
</listitem>
<listitem>
<para>
There is now better heap profiling support for
pinned objects (e.g. <literal>ByteStrings</literal>).
Previously we were completely ignoring them,
due to technical difficulties, but now count
all the pinned object space (including gaps between
pinned objects) as being type
<literal>ARR_WORDS</literal>. This isn't ideal, but
at least we do now account for the memory, and give
it the right type.
</para>
</listitem>
<listitem>
<para>
The quotes in an <literal>SCC</literal> pragma can
now be omitted if the SCC name is a valid Haskell
variable name, e.g.
<literal>{-# SCC my_function #-} expr</literal>.
</para>
</listitem>
<listitem>
<para>
It is now possible to use profiling when running on
multiple capabilities (i.e. when running with
<literal>+RTS -N</literal>). There is a new built-in
cost centre called <literal>IDLE</literal>, which
records the ticks of idle capabilities.
</para>
<para>
There are still some rough edges. In particular,
it is strongly recommended that you use the
<literal>-fno-prof-count-entries</literal> flag
or the program will run very slowly.
</para>
</listitem>
</itemizedlist>
</sect3>
<sect3>
<title>Event logging</title>
<itemizedlist>
<listitem>
<para>
TODO:
#define EVENT_SPARK_COUNTERS 34 /* (crt,dud,ovf,cnv,fiz,gcd,rem) */
#define EVENT_SPARK_CREATE 35 /* () */
#define EVENT_SPARK_DUD 36 /* () */
#define EVENT_SPARK_OVERFLOW 37 /* () */
#define EVENT_SPARK_RUN 38 /* () */
#define EVENT_SPARK_STEAL 39 /* (victim_cap) */
#define EVENT_SPARK_FIZZLE 40 /* () */
#define EVENT_SPARK_GC 41 /* () */
#define EVENT_INTERN_STRING 42 /* (string, id) {not used by ghc} */
TODO
d77df1caad3a5f833aac9275938a0675e1ee6aac
A new eventlog event containing 7 spark counters/statistics: sparks
created, dud, overflowed, converted, GC'd, fizzled and remaining.
These are maintained and logged separately for each capability.
We log them at startup, on each GC (minor and major) and on shutdown.
a5192d48e61a8ece69cddc43cc12625fcdcc56ec
add a new trace class for spark events
46b70749971341678c3e4d5cdb2b1ab1a13d039e
Move GC tracing into a separate trace class
Previously GC was included in the scheduler trace class. It can be
enabled specifically with +RTS -vg or -lg, though note that both -v
and -l on their own now default to a sensible set of trace classes,
currently: scheduler, gc and sparks.
084b64f22717b203b8d8c2ab7c057fb747e39593
Add new fully-accurate per-spark trace/eventlog events
Replaces the existing EVENT_RUN/STEAL_SPARK events with 7 new events
covering all stages of the spark lifcycle:
create, dud, overflow, run, steal, fizzle, gc
The sampled spark events are still available. There are now two event
classes for sparks, the sampled and the fully accurate. They can be
enabled/disabled independently. By default +RTS -l includes the sampled
but not full detail spark events. Use +RTS -lf-p to enable the detailed
'f' and disable the sampled 'p' spark.
b0935476b64b4818e4b5653c756c648623753fd3
Document the new +RTS -l flags in the +RTS --help output
8f4f29f655fdda443861152a24588fcaba29b168
Document the new +RTS -l options in the user guide
</para>
</listitem>
<listitem>
<para>
There is a new eventlog event
<literal>EVENT_WALL_CLOCK_TIME</literal>
which is used for matching the time of events
between different processes.
</para>
</listitem>
<listitem>
<para>
There is a new eventlog event
<literal>EVENT_THREAD_LABEL</literal>
which is emitted by the existing
<literal>GHC.Conc.labelThread</literal>
function. This allows tools such as
ThreadScope to use thread labels rather
than thread numbers.
</para>
</listitem>
<listitem>
<para>
TODO
From aaaaf67b2fca9bd9b0027c983bfc9f9255b2bce5 Mon Sep 17 00:00:00 2001
Subject: [PATCH 0695/1074] Add an RTS eventlog tracing class for user messages
</para>
</listitem>
<listitem>
<para>
On Windows, if the program name ends in
<literal>.exe</literal> then the
<literal>.exe</literal> is removed when making
the eventlog filename.
</para>
</listitem>
</itemizedlist>
</sect3>
<sect3>
<title>Runtime system</title>
<itemizedlist>
<listitem>
<para>
The restrictions on what
<literal>+RTS</literal>
flags are available by default has changed.
By default, you can now use the
<literal>-t</literal>,
<literal>-T</literal>,
<literal>-s</literal> and
<literal>-S</literal> RTS flags, provided you do not
give them a filename argument.
</para>
<para>
Additionally, when linking with
<literal>-threaded</literal>, you can now use the
<literal>+RTS -N</literal> flag without having
to link with <literal>-rtsopts</literal>.
Also, when linking with "developer" ways
(<literal>-debug</literal>,
<literal>-eventlog</literal> and
<literal>-prof</literal>) all the way-specific flags
are allowed; for example,
<literal>+RTS -h</literal> is allowed when linking
with <literal>-prof</literal>.
</para>
</listitem>
<listitem>
<para>
There is a new RTS flag
<literal>-T</literal>, which makes the RTS collect
statistics (memory usage, etc) but not give any
output. The new <literal>GHC.Stats</literal>
module in the <literal>base</literal> package
provides access to this data.
</para>
</listitem>
<listitem>
<para>
You can now give the RTS flag
<literal>-H</literal>
(without an argument) and the runtime system
will infer a sensible value to use. See
<xref linkend="rts-options-gc" /> for more details.
</para>
</listitem>
<listitem>
<para>
When using <literal>-no-hs-main</literal>
and starting the runtime system yourself,
if you wish to pass RTS flags then you will now
need to use the new <literal>hs_init_ghc</literal>
function. See <xref linkend="using-own-main" />
for details.
</para>
</listitem>
<listitem>
<para>
The runtime system now supports using
<literal>forkProcess</literal>
when running with multiple capabilities
(<literal>+RTS -N</literal>).
</para>
</listitem>
</itemizedlist>
</sect3>
<sect3>
<title>Build system</title>
<itemizedlist>
<listitem>
<para>
You can now build GHC with Alex 3.0.
</para>
</listitem>
<listitem>
<para>
TODO
On OS X, with XCode >= 4.2, GHC will use
gcc-4.2 rather than gcc.
</para>
</listitem>
<listitem>
<para>
There is now preliminary support for registerised
compilation on the ARM platform, using LLVM.
</para>
</listitem>
<listitem>
<para>
Dynamic libraries are now supported on OSX x86_64.
</para>
</listitem>
<listitem>
<para>
GHCi is now supported on kfreebsdgnu platforms.
</para>
</listitem>
<listitem>
<para>
GHC now recognises the s390x architecture.
</para>
</listitem>
</itemizedlist>
</sect3>
</sect2>
<sect2>
......@@ -320,7 +1360,7 @@
<itemizedlist>
<listitem>
<para>
Version number 0.5.1.0 (was 0.5.0.2)
Version number 0.5.0.3 (was 0.5.0.2)
It is now exposed by default.
</para>
</listitem>
......@@ -528,7 +1568,7 @@
<itemizedlist>
<listitem>
<para>
Version number 3.8.7.3 (was 3.8.7.1)
Version number 3.8.7.3 (was 3.8.7.2)
</para>
</listitem>
......@@ -654,8 +1694,7 @@
<listitem>
<para>
Both modules now export
<literal>reduceDoc</literal>
TODO: Why?
<literal>reduceDoc</literal>.
</para>
</listitem>
</itemizedlist>
......
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment