Commit 8d75c5b7 authored by simonmar's avatar simonmar
Browse files

[project @ 2000-05-11 16:07:37 by simonmar]

First cut at 4.07's release notes.
parent 70bad8db
<sect1 id="release-4-07">
<title>Release notes for version 4.07 (May 2000)</title>
<sect2>
<title>User-visible compiler changes</title>
<itemizedlist>
<listitem>
<para>New profiling subsystem, based on cost-centre stacks.
See <xref linkend="profiling">.</para>
</listitem>
<listitem>
<para>The x86 native code generator has been reworked
considerably, and now works reliably. Using the NCG rather
than compiling via C reduces compilation times by roughly a
half while having minimal effect on the run-time of the
compiled program (about 2-4% slower, worse for floating-point
intensive programs).</para>
<para>Add the <option>-fasm-x86</option> flag to GHC's command
line to enable it.</para>
</listitem>
<listitem>
<para>Implicit parameters.</para>
</listitem>
<listitem>
<para>Functional dependencies.</para>
</listitem>
<listitem>
<para>New <literal>DEPRECATED</literal> pragma for marking
outdated interfaces as deprecated.</para>
</listitem>
<listitem>
<para>New flag: <option>-fdump-minimal-imports</option>, which
dumps a file <literal>M.imports</literal> that contains the
(allegedly) minimal bunch of imports needed by the current
module.</para>
</listitem>
<listitem>
<para>New &ldquo;package&rdquo; system for libraries. See
<xref linkend="packages"> for the details.</para>
</listitem>
<listitem>
<para>The long-standing bug that caused some programs which
used <link linkend="trace"><literal>trace</literal></link> to
exit with a deadlock error has been fixed.</para>
</listitem>
<listitem>
<para>Trying to put into a full <link
linkend="sec-MVars"><literal>MVar</literal></link> will now
raise a <link
linkend="PutFullMVar"><literal>PutFullMVar</literal></link>
exception.</para>
</listitem>
<listitem>
<para>If a thread is about to be garbage collected, because it
is waiting on an <literal>MVar</literal> that no other thread
has access to, then it will now be sent the <link
linkend="BlockedOnDeadMVar"><literal>BlockedOnDeadMVar</literal></link>
exception.</para>
</listitem>
<listitem>
<para>A thread that is found to be blocked against itself
(i.e. is black holed) is now sent a
<literal>NonTermination</literal> exception.</para>
</listitem>
<listitem>
<para>Operations which may <emphasis>block</emphasis>, such as
<literal>takeMVar</literal>, <literal>raiseInThread</literal>,
and several I/O operations, may now receive asynchronous
exceptions even in the scope of a
<literal>blockAsyncExceptions</literal>. These are called
<firstterm>interruptible</firstterm> operations. See <xref
linkend="interruptible-operations"> for more details.</para>
</listitem>
<listitem>
<para>Result type signatures now work.</para>
</listitem>
<listitem>
<para>A truckload of bugfixes.</para>
</listitem>
</itemizedlist>
</sect2>
<sect2>
<title>User-visible library changes</title>
<itemizedlist>
<listitem>
<para>FFI.</para>
</listitem>
<listitem>
<para>HaXml, a library for parsing and generating XML, has
been added to the <literal>text</literal> package (<xref
linkend="sec-haxml">).</para>
</listitem>
<listitem>
<para>The <literal>QuickCheck</literal> library for performing
functional testing has been added to the
<literal>util</literal> package.</para>
</listitem>
<listitem>
<para>Two new experimental interfaces to arrays:
<literal>IArray</literal> for immutable arrays (<xref
linkend="sec-iarray">), and <literal>MArray</literal> for
mutable arrays (<xref linkend="sec-marray">).</para>
<para>Comments on these interfaces are welcome; eventually
we'd like them to replace <literal>ByteArray</literal>,
<literal>STArray</literal> and
<literal>IOArray</literal>.</para>
</listitem>
<listitem>
<para>New function: <link
linkend="tryTakeMVar"><literal>tryTakeMVar</literal></link>.</para>
</listitem>
<listitem>
<para><literal>MutableArray</literal> has gone away, use
<literal>STArray</literal> (see <xref linkend="sec-ST">)
instead.</para>
</listitem>
<listitem>
<para><literal>hPutBuf</literal>,
<literal>hPutBufBA</literal>, <literal>hGetBuf</literal>, and
<literal>hGetBufBA</literal>, have been renamed to
<literal>hPutBufFull</literal>,
<literal>hPutBufBAFull</literal>,
<literal>hGetBufFull</literal>, and
<literal>hGetBufBAFull</literal>. Functions with the old
names still exist, but have slightly different semantics. See
<xref linkend="sec-bulk-transfers"> for more details.</para>
</listitem>
</itemizedlist>
</sect2>
<sect2>
<title>Internal changes</title>
<itemizedlist>
<listitem>
<para><literal>Con</literal> is gone; the
<literal>CoreExpr</literal> type is simpler.</para>
</listitem>
<listitem>
<para><literal>NoRepLits</literal> have gone.</para>
</listitem>
<listitem>
<para>Better usage info in interface files, which means less
recompilation.</para>
</listitem>
<listitem>
<para>CCall primop is tidied up.</para>
</listitem>
<listitem>
<para>Constant folding now done by Rules.</para>
</listitem>
<listitem>
<para>Lots of hackery in the simplifier.</para>
</listitem>
<listitem>
<para>Improvements in CPR and strictness analysis.</para>
</listitem>
</itemizedlist>
</sect2>
</sect1>
<!-- Emacs stuff:
;;; Local Variables: ***
;;; mode: sgml ***
;;; sgml-parent-document: ("users_guide.sgml" "book" "chapter" "sect1") ***
;;; End: ***
-->
<!ENTITY license SYSTEM "license.sgml">
<!ENTITY intro SYSTEM "intro.sgml" >
<!ENTITY relnotes SYSTEM "4-06-notes.sgml" >
<!ENTITY relnotes SYSTEM "4-07-notes.sgml" >
<!ENTITY installing SYSTEM "installing.sgml" >
<!ENTITY using SYSTEM "using.sgml" >
<!ENTITY runtime SYSTEM "runtime_control.sgml" >
......
Markdown is supported
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