1 <?xml version="1.0" encoding="iso-8859-1"?>
2 <chapter id="bugs-and-infelicities">
3 <title>Known bugs and infelicities</title>
5 <sect1 id="vs-Haskell-defn">
6 <title>Haskell 98 vs. Glasgow Haskell: language non-compliance
9 <indexterm><primary>GHC vs the Haskell 98 language</primary></indexterm>
10 <indexterm><primary>Haskell 98 language vs GHC</primary></indexterm>
12 <para>This section lists Glasgow Haskell infelicities in its
13 implementation of Haskell 98. See also the “when things
14 go wrong” section (<xref linkend="wrong"/>) for information
15 about crashes, space leaks, and other undesirable phenomena.</para>
17 <para>The limitations here are listed in Haskell Report order
20 <sect2 id="haskell98-divergence">
21 <title>Divergence from Haskell 98</title>
24 <sect3 id="infelicities-lexical">
25 <title>Lexical syntax</title>
29 <para>The Haskell report specifies that programs may be
30 written using Unicode. GHC only accepts the ISO-8859-1
31 character set at the moment.</para>
35 <para>Certain lexical rules regarding qualified identifiers
36 are slightly different in GHC compared to the Haskell
38 <replaceable>module</replaceable><literal>.</literal><replaceable>reservedop</replaceable>,
39 such as <literal>M.\</literal>, GHC will interpret it as a
40 single qualified operator rather than the two lexemes
41 <literal>M</literal> and <literal>.\</literal>.</para>
46 <sect3 id="infelicities-syntax">
47 <title>Context-free syntax</title>
51 <para>GHC doesn't do fixity resolution in expressions during
52 parsing. For example, according to the Haskell report, the
53 following expression is legal Haskell:
55 let x = 42 in x == 42 == True</programlisting>
58 (let x = 42 in x == 42) == True</programlisting>
60 because according to the report, the <literal>let</literal>
61 expression <quote>extends as far to the right as
62 possible</quote>. Since it can't extend past the second
63 equals sign without causing a parse error
64 (<literal>==</literal> is non-fix), the
65 <literal>let</literal>-expression must terminate there. GHC
66 simply gobbles up the whole expression, parsing like this:
68 (let x = 42 in x == 42 == True)</programlisting>
70 The Haskell report is arguably wrong here, but nevertheless
71 it's a difference between GHC & Haskell 98.</para>
76 <sect3 id="infelicities-exprs-pats">
77 <title>Expressions and patterns</title>
79 <para>None known.</para>
82 <sect3 id="infelicities-decls">
83 <title>Declarations and bindings</title>
85 <para>None known.</para>
88 <sect3 id="infelicities-Modules">
89 <title>Module system and interface files</title>
91 <para>None known.</para>
94 <sect3 id="infelicities-numbers">
95 <title>Numbers, basic types, and built-in classes</title>
99 <term>Multiply-defined array elements—not checked:</term>
101 <para>This code fragment should
102 elicit a fatal error, but it does not:
105 main = print (array (1,1) [(1,2), (1,3)])</programlisting>
106 GHC's implementation of <literal>array</literal> takes the value of an
107 array slot from the last (index,value) pair in the list, and does no
108 checking for duplicates. The reason for this is efficiency, pure and simple.
116 <sect3 id="infelicities-Prelude">
117 <title>In <literal>Prelude</literal> support</title>
121 <term>Arbitrary-sized tuples</term>
123 <para>Tuples are currently limited to size 100. HOWEVER:
124 standard instances for tuples (<literal>Eq</literal>,
125 <literal>Ord</literal>, <literal>Bounded</literal>,
126 <literal>Ix</literal> <literal>Read</literal>, and
127 <literal>Show</literal>) are available
128 <emphasis>only</emphasis> up to 16-tuples.</para>
130 <para>This limitation is easily subvertible, so please ask
131 if you get stuck on it.</para>
136 <term><literal>Read</literal>ing integers</term>
138 <para>GHC's implementation of the
139 <literal>Read</literal> class for integral types accepts
140 hexadecimal and octal literals (the code in the Haskell
141 98 report doesn't). So, for example,
142 <programlisting>read "0xf00" :: Int</programlisting>
144 <para>A possible reason for this is that <literal>readLitChar</literal> accepts hex and
145 octal escapes, so it seems inconsistent not to do so for integers too.</para>
150 <term><literal>isAlpha</literal></term>
152 <para>The Haskell 98 definition of <literal>isAlpha</literal>
155 <programlisting>isAlpha c = isUpper c || isLower c</programlisting>
157 <para>GHC's implementation diverges from the Haskell 98
158 definition in the sense that Unicode alphabetic characters which
159 are neither upper nor lower case will still be identified as
160 alphabetic by <literal>isAlpha</literal>.</para>
167 <sect2 id="haskell98-undefined">
168 <title>GHC's interpretation of undefined behaviour in
169 Haskell 98</title>
171 <para>This section documents GHC's take on various issues that are
172 left undefined or implementation specific in Haskell 98.</para>
177 The <literal>Char</literal> type
178 <indexterm><primary><literal>Char</literal></primary><secondary>size of</secondary></indexterm>
181 <para>Following the ISO-10646 standard,
182 <literal>maxBound :: Char</literal> in GHC is
183 <literal>0x10FFFF</literal>.</para>
190 <indexterm><primary><literal>Int</literal></primary><secondary>size of</secondary></indexterm>
193 <para>In GHC the <literal>Int</literal> type follows the
194 size of an address on the host architecture; in other words
195 it holds 32 bits on a 32-bit machine, and 64-bits on a
196 64-bit machine.</para>
198 <para>Arithmetic on <literal>Int</literal> is unchecked for
199 overflow<indexterm><primary>overflow</primary><secondary><literal>Int</literal></secondary>
200 </indexterm>, so all operations on <literal>Int</literal> happen
202 2<superscript><replaceable>n</replaceable></superscript>
203 where <replaceable>n</replaceable> is the size in bits of
204 the <literal>Int</literal> type.</para>
206 <para>The <literal>fromInteger</literal><indexterm><primary><literal>fromInteger</literal></primary>
207 </indexterm>function (and hence
208 also <literal>fromIntegral</literal><indexterm><primary><literal>fromIntegral</literal></primary>
209 </indexterm>) is a special case when
210 converting to <literal>Int</literal>. The value of
211 <literal>fromIntegral x :: Int</literal> is given by taking
212 the lower <replaceable>n</replaceable> bits of <literal>(abs
213 x)</literal>, multiplied by the sign of <literal>x</literal>
214 (in 2's complement <replaceable>n</replaceable>-bit
215 arithmetic). This behaviour was chosen so that for example
216 writing <literal>0xffffffff :: Int</literal> preserves the
217 bit-pattern in the resulting <literal>Int</literal>.</para>
220 <para>Negative literals, such as <literal>-3</literal>, are
221 specified by (a careful reading of) the Haskell Report as
222 meaning <literal>Prelude.negate (Prelude.fromInteger 3)</literal>.
223 So <literal>-2147483648</literal> means <literal>negate (fromInteger 2147483648)</literal>.
224 Since <literal>fromInteger</literal> takes the lower 32 bits of the representation,
225 <literal>fromInteger (2147483648::Integer)</literal>, computed at type <literal>Int</literal> is
226 <literal>-2147483648::Int</literal>. The <literal>negate</literal> operation then
227 overflows, but it is unchecked, so <literal>negate (-2147483648::Int)</literal> is just
228 <literal>-2147483648</literal>. In short, one can write <literal>minBound::Int</literal> as
229 a literal with the expected meaning (but that is not in general guaranteed.
232 <para>The <literal>fromIntegral</literal> function also
233 preserves bit-patterns when converting between the sized
234 integral types (<literal>Int8</literal>,
235 <literal>Int16</literal>, <literal>Int32</literal>,
236 <literal>Int64</literal> and the unsigned
237 <literal>Word</literal> variants), see the modules
238 <literal>Data.Int</literal> and <literal>Data.Word</literal>
239 in the library documentation.</para>
244 <term>Unchecked float arithmetic</term>
246 <para>Operations on <literal>Float</literal> and
247 <literal>Double</literal> numbers are
248 <emphasis>unchecked</emphasis> for overflow, underflow, and
249 other sad occurrences. (note, however that some
250 architectures trap floating-point overflow and
251 loss-of-precision and report a floating-point exception,
252 probably terminating the
253 program)<indexterm><primary>floating-point
254 exceptions</primary></indexterm>.</para>
264 <title>Known bugs or infelicities</title>
266 <para>In addition to the divergences from the Haskell 98 standard
267 listed above, GHC has the following known bugs or
270 <sect2 id="bugs-ghc">
271 <title>Bugs in GHC</title>
275 <para> GHC can warn about non-exhaustive or overlapping
276 patterns (see <xref linkend="options-sanity"/>), and usually
277 does so correctly. But not always. It gets confused by
278 string patterns, and by guards, and can then emit bogus
279 warnings. The entire overlap-check code needs an overhaul
284 <para>GHC does not allow you to have a data type with a context
285 that mentions type variables that are not data type parameters.
288 data C a b => T a = MkT a
290 so that <literal>MkT</literal>'s type is
292 MkT :: forall a b. C a b => a -> T a
294 In principle, with a suitable class declaration with a functional dependency,
295 it's possible that this type is not ambiguous; but GHC nevertheless rejects
296 it. The type variables mentioned in the context of the data type declaration must
297 be among the type parameters of the data type.</para>
301 <para>GHC's inliner can be persuaded into non-termination
302 using the standard way to encode recursion via a data type:</para>
304 data U = MkU (U -> Bool)
307 russel u@(MkU p) = not $ p u
310 x = russel (MkU russel)
313 <para>We have never found another class of programs, other
314 than this contrived one, that makes GHC diverge, and fixing
315 the problem would impose an extra overhead on every
316 compilation. So the bug remains un-fixed. There is more
318 url="http://research.microsoft.com/~simonpj/Papers/inlining">
319 Secrets of the GHC inliner</ulink>.</para>
324 <sect2 id="bugs-ghci">
325 <title>Bugs in GHCi (the interactive GHC)</title>
328 <para>GHCi does not respect the <literal>default</literal>
329 declaration in the module whose scope you are in. Instead,
330 for expressions typed at the command line, you always get the
331 default default-type behaviour; that is,
332 <literal>default(Int,Double)</literal>.</para>
334 <para>It would be better for GHCi to record what the default
335 settings in each module are, and use those of the 'current'
336 module (whatever that is).</para>
340 <para>GHCi does not keep careful track of what instance
341 declarations are 'in scope' if they come from other packages.
342 Instead, all instance declarations that GHC has seen in other
343 packages are all in scope everywhere, whether or not the
344 module from that package is used by the command-line
349 <para>On Windows, there's a GNU ld/BFD bug
350 whereby it emits bogus PE object files that have more than
351 0xffff relocations. When GHCi tries to load a package affected by this
352 bug, you get an error message of the form
354 Loading package javavm ... linking ... WARNING: Overflown relocation field (# relocs found: 30765)
356 The last time we looked, this bug still
357 wasn't fixed in the BFD codebase, and there wasn't any
358 noticeable interest in fixing it when we reported the bug
361 <para>The workaround is to split up the .o files that make up
362 your package into two or more .o's, along the lines of
363 how the "base" package does it.</para>
372 ;;; Local Variables: ***
374 ;;; sgml-parent-document: ("users_guide.xml" "book" "chapter") ***