1 <!-- -*- mode: SGML; sgml-parent-document: ("user-manual.sgml" "BOOK") -*- -->
2 <chapter id="beyond-ansi"><title>Beyond The &ANSI; Standard</>
4 <para>&SBCL; is mostly an implementation of the &ANSI; standard for
5 Common Lisp. However, there's some important behavior which extends
6 or clarifies the standard, and various behavior which outright
10 <sect1 id="non-conformance"><title>Non-Conformance With The &ANSI; Standard</>
13 Essentially every type of non-conformance is considered a bug.
14 (The exceptions involve internal inconsistencies in the standard.)
15 In &SBCL; 0.7.6, the master record of known bugs is in
16 the <filename>BUGS</> file in the distribution.
17 Some highlight information about bugs may also be found in the
18 manual page. The recommended way to report bugs is through the sbcl-help or
19 sbcl-devel mailings lists.
24 <sect1 id="idiosyncrasies"><title>Idiosyncrasies</>
26 <para>The information in this section describes some of the ways
27 that &SBCL; deals with choices that the &ANSI; standard
28 leaves to the implementation.</para>
30 <para>Declarations are generally treated as assertions. This general
31 principle, and its implications, and the bugs which still keep the
32 compiler from quite satisfying this principle, are discussed in the
33 <link linkend="compiler">chapter on the compiler</link>.</para>
35 <para>&SBCL; is essentially a compiler-only implementation of
36 &CommonLisp;. That is, for all but a few special cases,
37 <function>eval</> creates a
38 lambda expression, calls <function>compile</> on the lambda
39 expression to create a compiled function, and then calls
40 <function>funcall</> on the resulting function object. This
41 is explicitly allowed by the &ANSI; standard, but leads to some
42 oddities, e.g. collapsing <function>functionp</> and
43 <function>compiled-function-p</> into the same predicate.</para>
45 <para>&SBCL; is quite strict about ANSI's definition of
46 <function>defconstant</>. ANSI says that doing <function>defconstant</>
47 of the same symbol more than once is undefined unless the new value
48 is <function>eql</> to the old value. Conforming to this specification
49 is a nuisance when the "constant" value is only constant under some
50 weaker test like <function>string=</> or <function>equal</>. It's
51 especially annoying because, in &SBCL;, <function>defconstant</> takes effect
52 not only at load time but also at compile time, so that just
53 compiling and loading reasonable code like
54 <programlisting>(defconstant +foobyte+ '(1 4))</>
55 runs into this undefined behavior. Many
56 implementations of Common Lisp try to help the programmer around
57 this annoyance by silently accepting the undefined code and
58 trying to do what the programmer probably meant. &SBCL; instead
59 treats the undefined behavior as an error. Often
60 such code can be rewritten
61 in portable &ANSI; Common Lisp which has the desired behavior.
62 E.g., the code above can be given an exactly defined meaning by replacing
63 <function>defconstant</> either with <function>defparameter</> or
64 with a customized macro which does the right thing, possibly along the
65 lines of the <function>defconstant-eqx</> macro used internally in the
66 implementation of &SBCL; itself. In circumstances where this is not
67 appropriate, the programmer can handle the condition type
68 <errortype>sb-ext:defconstant-uneql</errortype>, and choose either the
69 <action>continue</action> or <action>abort</action> restart as
72 <para>&SBCL; gives style warnings about various kinds of perfectly
75 <listitem><para><function>defmethod</> without
76 <function>defgeneric</></para></listitem>
77 <listitem><para>multiple <function>defun</>s of the same
78 symbol</para></listitem>
79 <listitem><para>special variables not named in the conventional
80 <varname>*foo*</> style, and lexical variables unconventionally named
81 in the <varname>*foo*</> style</para></listitem>
83 This causes friction with people
84 who point out that other ways of organizing code (especially
85 avoiding the use of <function>defgeneric</>)
86 are just as aesthetically stylish.
87 However, these warnings should be read not
88 as "warning, bad aesthetics detected, you have no style" but
89 "warning, this style keeps the compiler from understanding
90 the code as well as you might like." That is,
91 unless the compiler warns about such conditions, there's no
92 way for the compiler to warn
93 about some programming errors which would otherwise be
94 easy to overlook. (related bug: The warning about
95 multiple <function>defun</>s is pointlessly annoying when you compile
96 and then load a function containing <function>defun</> wrapped
97 in <function>eval-when</>, and ideally should be suppressed in
98 that case, but still isn't as of &SBCL; 0.7.6.)</para>
102 <sect1 id="extensions"><title>Extensions</>
104 <para>&SBCL; is derived from &CMUCL;, which implements many extensions
105 to the &ANSI; standard. &SBCL; doesn't support as many extensions as
106 &CMUCL;, but it still has quite a few.</para>
108 <sect2><title>Things Which Might Be In The Next &ANSI; Standard</>
110 <para>&SBCL; provides extensive support for
111 calling external C code, described
112 <link linkend="ffi">in its own chapter</link>.</para>
114 <para>&SBCL; provides additional garbage collection functionality not
115 specified by &ANSI;. Weak pointers allow references to objects to be
116 maintained without keeping them from being GCed. And "finalization"
117 hooks are available to cause code to be executed when an object has been
118 GCed.</para> <!-- FIXME: Actually documenting these would be good.:-| -->
120 <para>&SBCL; supports Gray streams, user-overloadable CLOS classes
121 whose instances can be used as Lisp streams (e.g. passed as the
122 first argument to <function>format</>). Additionally, the
123 bundled contrib module <interface>sb-simple-streams</interface>
124 implements a subset of the Franz Allegro simple-streams proposal.</para>
126 <para>&SBCL; supports a MetaObject Protocol which is intended to be
127 compatible with &AMOP;; present exceptions to this (as distinct from
130 <listitem><para>the abstract <classname>metaobject</> class is not
131 present in the class hierarchy;</para></listitem>
132 <listitem><para>the <classname>standard-object</> and
133 <classname>funcallable-standard-object</> classes are
134 disjoint;</para></listitem>
135 <listitem><para><function>compute-effective-method</> only returns
136 one value, not two;</para></listitem>
137 <listitem><para>the system-supplied <property>:around</> method for
138 <function>compute-slots</> specialized on
139 <classname>funcallable-standard-class</> does not respect the
140 requested order from a user-supplied primary method.
145 <sect2><title>Threading (a.k.a Multiprocessing)</>
147 <para>&SBCL; (as of version 0.8.3, on Linux x86 only) supports a
148 fairly low-level threading interface that maps onto the host operating
149 system's concept of threads or lightweight processes.
151 <sect3><title>Lisp-level view</title>
153 <para>A rudimentary interface to creating and managing multiple threads
154 can be found in the <literal>sb-thread</literal> package. This is
155 intended for public consumption, so look at the exported symbols and
156 their documentation strings.
158 <para>Dynamic bindings to symbols are per-thread. Signal handlers
161 <para><function>sb-ext:quit</function> exits the current thread, not
162 necessarily the whole environment. The environment will be shut down
163 when the last thread exits.
165 <para>Threads arbitrate between themselves for the user's attention.
166 A thread may be in one of three notional states: foreground,
167 background, or stopped. When a background process attempts to print a
168 repl prompt or to enter the debugger, it will stop and print a message
169 saying that it has stopped. The user at his leisure may switch to
170 that thread to find out what it needs. If a background thread enters
171 the debugger, selecting any restart will put it back into the
172 background before it resumes.
174 <para>If the user has multiple views onto the same Lisp image (for
175 example, using multiple terminals, or a windowing system, or network
176 access) they are typically set up as multiple `sessions' such that each
177 view has its own collection of foreground/background/stopped threads.
178 <function>sb-thread:make-listener-thread</function> can be used to
179 start a new thread in its own `session'.
181 <para>Mutexes and condition variables are available for
182 managing access to shared data: see
186 <programlisting>(apropos "mutex" :sb-thread)</programlisting>
188 <programlisting>(apropos "condition" :sb-thread)</programlisting>
189 <listitem> <para>and the <structname>waitqueue</structname> structure
194 and poke around in their documentation strings.
196 <sect3><title>Implementation (Linux x86)</title>
198 <para>On Linux x86, this is implemented using
199 <function>clone()</function> and does not involve pthreads. This is
200 not because there is anything wrong with pthreads <emphasis>per
201 se</emphasis>, but there is plenty wrong (from our perspective) with
202 LinuxThreads. &SBCL; threads are mapped 1:1 onto Linux tasks which
203 share a VM but nothing else - each has its own process id and can be
204 seen in e.g. <command>ps</command> output.
206 <para>Per-thread local bindings for special variables is achieved
207 using the %fs segment register to point to a per-thread storage area.
208 This may cause interesting results if you link to foreign code that
209 expects threading or creates new threads, and the thread library in
210 question uses %fs in an incompatible way.
212 <para>Threads waiting on queues (e.g. for locks or condition
213 variables) are put to sleep using <function>sigtimedwait()</function>
214 and woken with SIGCONT.
216 <para>&SBCL; at present will alway have at least two tasks running as
217 seen from Linux: when the first process has done startup
218 initialization (mapping files in place, installing signal handlers
219 etc) it creates a new thread to run the Lisp startup and initial
220 listener. The original thread stays around to reap dead subthreads
221 and deallocate their resources (e.g. stacks) when they exit.
223 <para>Garbage collection is done with the existing Conservative
224 Generational GC. Allocation is done in small (typically 8k) regions :
225 each thread has its own region so this involves no stopping. However,
226 when a region fills, a lock must be obtained while another is
227 allocated, and when a collection is required, all processes are
228 stopped. This is achieved by sending them signals, which may make for
229 interesting behaviour if they are interrupted in system calls. The
230 streams interface is believed to handle the required system call
231 restarting correctly, but this may be a consideration when making
232 other blocking calls e.g. from foreign library code.
234 <para>Large amounts of the &SBCL; library have not been inspected for
235 thread-safety. Some of the obviously unsafe areas have large locks
236 around them, so compilation and fasl loading, for example, cannot be
237 parallelized. Work is ongoing in this area.
239 <para>A new thread by default is created in the same POSIX process
240 group and session as the thread it was created by. This has an impact
241 on keyboard interrupt handling: pressing your terminal's intr key
242 (typically Control-C) will interrupt all processes in the foreground
243 process group, including Lisp threads that &SBCL; considers to be
244 notionally `background'. This is undesirable, so background threads
245 are set to ignore the SIGINT signal. Arbitration for the input stream
246 is managed by locking on <varname>sb-thread::*session-lock*</varname>
248 <para>A thread can be created in a new Lisp 'session' (new terminal or
249 window) using <function>sb-thread:make-listener-thread</function>.
250 These sessions map directly onto POSIX sessions, so that pressing
251 Control-C in the wrong window will not interrupt them - this has been
252 found to be embarrassing.
254 <sect2><title>Support For Unix</>
256 <para>The UNIX command line can be read from the variable
257 <varname>sb-ext:*posix-argv*</>. The UNIX environment can be queried with the
258 <function>sb-ext:posix-getenv</> function.</para>
260 <para>The &SBCL; system can be terminated with <function>sb-ext:quit</>,
261 optionally returning a specified numeric value to the calling Unix
262 process. The normal Unix idiom of terminating on end of file on input
263 is also supported.</para>
267 <sect2><title>Customization Hooks for Users</title>
269 <para>The behaviour of <function>require</function> when called with only
270 one argument is implementation-defined. In &SBCL; it calls functions
271 on the user-settable list <varname>sb-ext:*module-provider-functions*</varname>
272 - see the <function>require</function> documentation string for details.
274 <para>The toplevel repl prompt may be customized, and the function
275 that reads user input may be replaced completely. <!-- FIXME but I
276 don't currently remember how -->
278 <sect2><title>Tools To Help Developers</title>
280 <para>&SBCL; provides a profiler and other extensions to the &ANSI;
281 <function>trace</> facility. See the online function documentation for
282 <function>trace</> for more information.</para>
284 <para>The debugger supports a number of options. Its documentation is
285 accessed by typing <userinput>help</> at the debugger prompt.</para>
287 A true debugger section in the manual would be good. Start
288 with CMU CL's debugger section, but remember:
289 * no QUIT command (TOPLEVEL restart instead)
290 * no GO command (CONTINUE restart instead)
291 * Limitations of the x86 port of the debugger should be
292 documented or fixed where possible.
293 * Discuss TRACE and its unification with PROFILE. -->
295 <para>Documentation for <function>inspect</> is accessed by typing
296 <userinput>help</> at the <function>inspect</> prompt.</para>
300 <sect2><title>Interface To Low-Level &SBCL; Implementation</title>
302 <para>&SBCL; has the ability to save its state as a file for later
303 execution. This functionality is important for its bootstrapping
304 process, and is also provided as an extension to the user See the
305 documentation for <function>sb-ext:save-lisp-and-die</> for more
308 <note><para>&SBCL; has inherited from &CMUCL; various hooks to allow
309 the user to tweak and monitor the garbage collection process. These
310 are somewhat stale code, and their interface might need to be cleaned
311 up. If you have urgent need of them, look at the code in
312 <filename>src/code/gc.lisp</filename> and bring it up on the
313 developers' mailing list.</para></note>
315 <note><para>&SBCL; has various hooks inherited from &CMUCL;, like
316 <function>sb-ext:float-denormalized-p</>, to allow a program to take
317 advantage of &IEEE; floating point arithmetic properties which aren't
318 conveniently or efficiently expressible using the &ANSI; standard. These
319 look good, and their interface looks good, but &IEEE; support is
320 slightly broken due to a stupid decision to remove some support for
321 infinities (because it wasn't in the &ANSI; spec and it didn't occur to
322 me that it was in the &IEEE; spec). If you need this stuff, take a look
323 at the code and bring it up on the developers' mailing
328 <sect2><title>Efficiency Hacks</title>
330 <para>The <function>sb-ext:purify</function> function causes &SBCL;
331 first to collect all garbage, then to mark all uncollected objects as
332 permanent, never again attempting to collect them as garbage. This can
333 cause a large increase in efficiency when using a primitive garbage
334 collector, or a more moderate increase in efficiency when using a more
335 sophisticated garbage collector which is well suited to the program's
336 memory usage pattern. It also allows permanent code to be frozen at
337 fixed addresses, a precondition for using copy-on-write to share code
338 between multiple Lisp processes. is less important with modern
339 generational garbage collectors. </para>
341 <para>The <function>sb-ext:truly-the</> declares the type of the
342 result of the operations, producing its argument; the declaration is
343 not checked. In short: don't use it.</para>
345 <para>The <function>sb-ext:freeze-type</> declaration declares that a
346 type will never change, which can make type testing
347 (<function>typep</>, etc.) more efficient for structure types.</para>
349 <para>The <function>sb-ext:constant-function</> declaration specifies
350 that a function will always return the same value for the same
351 arguments, which may allow the compiler to optimize calls
352 to it. This is appropriate for functions like <function>sqrt</>, but
353 is <emphasis>not</> appropriate for functions like <function>aref</>,
354 which can change their return values when the underlying data are
356 <!-- FIXME: This declaration does not seem to be supported in the --
357 -- current compiler. -->