1 <Chapter id="bugs-and-infelicities">
2 <title>Known bugs and infelicities</title>
4 <sect1 id="vs-Haskell-defn">
5 <title>Haskell 98 vs. Glasgow Haskell: language non-compliance
8 <indexterm><primary>GHC vs the Haskell 98 language</primary></indexterm>
9 <indexterm><primary>Haskell 98 language vs GHC</primary></indexterm>
11 <para>This section lists Glasgow Haskell infelicities in its
12 implementation of Haskell 98. See also the “when things
13 go wrong” section (<XRef LinkEnd="wrong">) for information
14 about crashes, space leaks, and other undesirable phenomena.</para>
16 <para>The limitations here are listed in Haskell Report order
19 <sect2 id="haskell98-divergence">
20 <title>Divergence from Haskell 98</title>
23 <sect3 id="infelicities-lexical">
24 <title>Lexical syntax</title>
28 <para>The Haskell report specifies that programs may be
29 written using Unicode. GHC only accepts the ISO-8859-1
30 character set at the moment.</para>
34 <para>Certain lexical rules regarding qualified identifiers
35 are slightly different in GHC compared to the Haskell
37 <replaceable>module</replaceable><literal>.</literal><replaceable>reservedop</replaceable>,
38 such as <literal>M.\</literal>, GHC will interpret it as a
39 single qualified operator rather than the two lexemes
40 <literal>M</literal> and <literal>.\</literal>.</para>
45 <sect3 id="infelicities-syntax">
46 <title>Context-free syntax</title>
50 <para>GHC doesn't do fixity resolution in expressions during
51 parsing. For example, according to the Haskell report, the
52 following expression is legal Haskell:
54 let x = 42 in x == 42 == True</programlisting>
57 (let x = 42 in x == 42) == True</programlisting>
59 because according to the report, the <literal>let</literal>
60 expression <quote>extends as far to the right as
61 possible</quote>. Since it can't extend past the second
62 equals sign without causing a parse error
63 (<literal>==</literal> is non-fix), the
64 <literal>let</literal>-expression must terminate there. GHC
65 simply gobbles up the whole expression, parsing like this:
67 (let x = 42 in x == 42 == True)</programlisting>
69 The Haskell report is arguably wrong here, but nevertheless
70 it's a difference between GHC & Haskell 98.</para>
75 <sect3 id="infelicities-exprs-pats">
76 <title>Expressions and patterns</title>
78 <para>None known.</para>
82 <sect3 id="infelicities-decls">
83 <title>Declarations and bindings</title>
85 <para>None known.</para>
89 <sect3 id="infelicities-Modules">
90 <title>Module system and interface files</title>
94 <term><literal>Main</literal> module</term>
96 <para>GHC interprets the module header
97 <programlisting>module Main where</programlisting>
99 <programlisting>module Main (main) where</programlisting>
102 <para>This change allows GHC to optimise slightly more
103 aggresively inside the <literal>Main</literal>
106 <para>You are highly unlikely to notice the difference, since
107 importing <literal>Main</literal> is very rare (it would
108 introduce a recursive module dependency, so doing it by
109 accident is unlikely too).</para>
116 <sect3 id="infelicities-numbers">
117 <title>Numbers, basic types, and built-in classes</title>
121 <term>Multiply-defined array elements—not checked:</term>
123 <para>This code fragment <emphasis>should</emphasis>
124 elicit a fatal error, but it does not:
127 main = print (array (1,1) [(1,2), (1,3)])</programlisting>
136 <sect3 id="infelicities-Prelude">
137 <title>In <literal>Prelude</literal> support</title>
141 <term>The <literal>Char</literal> type</term>
142 <indexterm><primary><literal>Char</literal></primary><secondary>size
143 of</secondary></indexterm>
145 <para>The Haskell report says that the
146 <literal>Char</literal> type holds 16 bits. GHC follows
147 the ISO-10646 standard a little more closely:
148 <literal>maxBound :: Char</literal> in GHC is
149 <literal>0x10FFFF</literal>.</para>
154 <term>Arbitrary-sized tuples</term>
156 <para>Tuples are currently limited to size 100. HOWEVER:
157 standard instances for tuples (<literal>Eq</literal>,
158 <literal>Ord</literal>, <literal>Bounded</literal>,
159 <literal>Ix</literal> <literal>Read</literal>, and
160 <literal>Show</literal>) are available
161 <emphasis>only</emphasis> up to 16-tuples.</para>
163 <para>This limitation is easily subvertible, so please ask
164 if you get stuck on it.</para>
169 <term><literal>Read</literal>ing integers</term>
171 <para>GHC's implementation of the
172 <literal>Read</literal> class for integral types accepts
173 hexadeciaml and octal literals (the code in the Haskell
174 98 report doesn't). So, for example,
175 <programlisting>read "0xf00" :: Int</programlisting>
177 <para>A possible reason for this is that <literal>readLitChar</literal> accepts hex and
178 octal escapes, so it seems inconsistent not to do so for integers too.</para>
185 <sect2 id="haskell98-undefined">
186 <title>GHC's interpretation of undefined behaviour in
187 Haskell 98</title>
189 <para>This section documents GHC's take on various issues that are
190 left undefined or implementation specific in Haskell 98.</para>
194 <term>Sized integral types</term>
195 <indexterm><primary><literal>Int</literal></primary><secondary>size of</secondary>
199 <para>In GHC the <literal>Int</literal> type follows the
200 size of an address on the host architecture; in other words
201 it holds 32 bits on a 32-bit machine, and 64-bits on a
202 64-bit machine.</para>
204 <para>Arithmetic on <literal>Int</literal> is unchecked for
205 overflow<indexterm><primary>overflow</primary><secondary><literal>Int</literal></secondary>
206 </indexterm>, so all operations on <literal>Int</literal> happen
208 2<superscript><replaceable>n</replaceable></superscript>
209 where <replaceable>n</replaceable> is the size in bits of
210 the <literal>Int</literal> type.</para>
212 <para>The <literal>fromInteger</literal><indexterm><primary><literal>fromInteger</literal></primary>
213 </indexterm>function (and hence
214 also <literal>fromIntegral</literal><indexterm><primary><literal>fromIntegral</literal></primary>
215 </indexterm>) is a special case when
216 converting to <literal>Int</literal>. The value of
217 <literal>fromIntegral x :: Int</literal> is given by taking
218 the lower <replaceable>n</replaceable> bits of <literal>(abs
219 x)</literal>, multiplied by the sign of <literal>x</literal>
220 (in 2's complement <replaceable>n</replaceable>-bit
221 arithmetic). This behaviour was chosen so that for example
222 writing <literal>0xffffffff :: Int</literal> preserves the
223 bit-pattern in the resulting <literal>Int</literal>.</para>
226 <para>Negative literals, such as <literal>-3</literal>, are
227 specified by (a careful reading of) the Haskell Report as
228 meaning <literal>Prelude.negate (Prelude.fromInteger 3)</literal>.
229 So <literal>-2147483648</literal> means <literal>negate (fromInteger 2147483648)</literal>.
230 Since <literal>fromInteger</literal> takes the lower 32 bits of the representation,
231 <literal>fromInteger (2147483648::Integer)</literal>, computed at type <literal>Int</literal> is
232 <literal>-2147483648::Int</literal>. The <literal>negate</literal> operation then
233 overflows, but it is unchecked, so <literal>negate (-2147483648::Int)</literal> is just
234 <literal>-2147483648</literal>. In short, one can write <literal>minBound::Int</literal> as
235 a literal with the expected meaning (but that is not in general guaranteed.
238 <para>The <literal>fromIntegral</literal> function also
239 preserves bit-patterns when converting between the sized
240 integral types (<literal>Int8</literal>,
241 <literal>Int16</literal>, <literal>Int32</literal>,
242 <literal>Int64</literal> and the unsigned
243 <literal>Word</literal> variants), see the modules
244 <literal>Data.Int</literal> and <literal>Data.Word</literal>
245 in the library documentation.</para>
250 <term>Unchecked float arithmetic</term>
252 <para>Operations on <literal>Float</literal> and
253 <literal>Double</literal> numbers are
254 <emphasis>unchecked</emphasis> for overflow, underflow, and
255 other sad occurrences. (note, however that some
256 architectures trap floating-point overflow and
257 loss-of-precision and report a floating-point exception,
258 probably terminating the
259 program)<indexterm><primary>floating-point
260 exceptions</primary></indexterm>.</para>
270 <title>Known bugs or infelicities</title>
272 <para>In addition to the divergences from the Haskell 98 standard
273 listed above, GHC has the following known bugs or
278 <para> GHC can warn about non-exhaustive or overlapping
279 patterns (see <xref linkend="options-sanity">, and usually
280 does so correctly. But not always. It gets confused by
281 string patterns, and by guards, and can then emit bogus
282 warnings. The entire overlap-check code needs an overhaul
287 <para>Dangers with multiple <literal>Main</literal>
290 <para>GHC does not insist that module <literal>Main</literal>
291 lives in a file called <filename>Main.hs</filename>. This is
292 useful if you want multiple versions of
293 <literal>Main</literal>. But there's a danger: when compiling
294 module <literal>Main</literal> (regardless of what file it
295 comes from), GHC looks for the interface
296 <filename>Main.hi</filename>; it uses this to get version
297 information from the last time it recompiled
298 <literal>Main</literal>. The trouble is that this
299 <filename>Main.hi</filename> may not correspond to the source
300 file being compiled.</para>
302 <para>Solution: remove <filename>Main.hi</filename> first. A
303 better solution would be for GHC to record the source-file
304 filename in the interface file, or even an MD5 checksum.
309 <para>GHCi does not respect the <literal>default</literal>
310 declaration in the module whose scope you are in. Instead,
311 for expressions typed at the command line, you always get the
312 default default-type behaviour; that is,
313 <literal>default(Int,Double)</literal>.</para>
315 <para>It would be better for GHCi to record what the default
316 settings in each module are, and use those of the 'current'
317 module (whatever that is).</para>
321 <para>GHCi does not keep careful track of what instance
322 declarations are 'in scope' if they come from other packages.
323 Instead, all instance declarations that GHC has seen in other
324 packages are all in scope everywhere, whether or not the
325 module from that package is used by the command-line
330 <para>GHC's inliner can be persuaded into non-termination
331 using the standard way to encode recursion via a data type:</para>
333 data U = MkU (U -> Bool)
336 russel u@(MkU p) = not $ p u
339 x = russel (MkU russel)
342 <para>We have never found another class of programs, other
343 than this contrived one, that makes GHC diverge, and fixing
344 the problem would impose an extra overhead on every
345 compilation. So the bug remains un-fixed. There is more
347 url="http://research.microsoft.com/~simonpj/Papers/inlining">
348 Secrets of the GHC inliner</ulink>.</para>
356 ;;; Local Variables: ***
358 ;;; sgml-parent-document: ("users_guide.sgml" "book" "chapter" "sect1") ***