This is a list of frequently asked questions about &clisp; on the CLISP mailing lists and the USENET newsgroup comp.lang.lisp. All the legitimate technical question are addressed in the &clisp; documentation (, , ), and for such questions this list provides a link into the docs. The frequently asked political questions are answered here in full detail (meaning that no further explanations of the issues could be provided). Please submit more questions (and answers!) to clisp-list. What is FAQ fine? We levy a fine of 10 zorkmids for asking a question that is answered with a link to this FAQ document. We further levy a fine of 1 zorkmid for asking a question that is answered with a link to the &clisp; manual. The fines are payable to the person who answered the questions. The definition of a zorkmid is left to the sole discretion of the payer. This should ¬-e; discourage you from asking questions, but rather encourage you to read the manual and answer questions from other users. The official &clisp; documentation sucks - is anything better available? As with all generic complaints, the answer to this one is PTCPatches Thoughtfully Considered. Additionally, the nightly builds of the &clisp; implementation notes documenting the current &hg; default branch are available at . It contains both the documentation for the new features and the general improvements in the documentation. License - why &gnu; &gpl;? Because &clisp; uses &readline;. Note that this does ¬-e; necessarily prevent you from distributing your proprietary products based on &clisp;. See Note in COPYRIGHT and . What about &ansi-cl; compliance? &clisp; purports to conform to the &ansi-cl; specification, so all deviations in &sh-prompt; &clisp-cmd; &opt-ansi; from the &ansi-cl; are bugs and are not (yet) fixed only due to lack of resources. On the other hand, some decisions made by the ANSI X3J13 committee were not as justified from the technical point of view as were most of them, and some of those questionable decisions were made &after-e; the alternative behavior has already been implemented in &clisp;. The &clisp; developers took pains to modify &clisp; to unconditionally comply with the &ansi-cl; specification in all cases except for a handful of situations where they believed that the committee had made a mistake, in which cases the committee behavior is still optionally available. &clisp; does not start in the ansi mode by default for historical reasons and this is not about to change. Dumping an image or passing a command line argument are easy enough. How do I ask for help? &clisp;Type &help; and hit &enter;. Human Politely - please refer to &netiquette;. If you have a question about &clisp;, you have the following options (listed in the order of decreasing audience size): USENET group comp.lang.lisp This is the right place to ask all general Lisp questions, such as "how do I access a character in a string?" &clisp; User Mailing List clisp-list AKA This is the right place to ask user-level &clisp;-specific questions, such as "how do I save my &clisp; &mem-image;?". See also . &clisp; Developer Mailing List clisp-devel AKA This is the right place to discuss &clisp; internals, problems with the current development sources, submit patches, offer help. Individual &clisp; developers This is never the right thing to do, unless you want to hire this developer to do &clisp; consulting (commercial support, custom enhancements etc). This is actually for your own benefit: the &clisp; developers are very busy, they might get weeks to answer your question, and some other user may be able to help you in the meantine; as well as for the benefit of the other &clisp; users, since the mailing lists are publicly archived (you are encouraged to search them before asking). Please do ¬-e; copy your messages to the individual developers. To avoid spam, we require that you subscribe before you can post to a mailing list. Alternatively, you can post to the &clisp; web forum and your message will be forwarded to the clisp-list mailing list. If you read a list on Gmane and do not want to receive it also by e-mail, you can subscribe to it using the aforementioned web interface and then disable mail delivery. Which mailing lists should I subscribe to? Cross-posting in the &clisp; mailing lists is very actively discouraged and is virtually non-existent, thus you can subscribe to all mailing lists that are relevant to you without getting duplicate messages: clisp-announce Extremely low-level moderated list, you should definitely subscribe to this list if you have any interest in &cl; or &clisp; whatsoever. clisp-list Subscribe to this list of you use &clisp; and want to ask (and answer!) questions. If you subscribe to this list, you probably want to subscribe to clisp-announce too. clisp-devel Subscribe to this list if you want to influence &clisp; developments; &hg; commit notes go here (as a daily digest). If you subscribe to this list, you probably want to subscribe to clisp-list &and-e; clisp-announce too. Why is my mail to a mailing list rejected? &clisp; mailing lists get a lot of spam, so the maintainers have to take care to protect the users. If you get a note that your message is held for moderator's approval, you can safely assume that it went to /dev/null and try again, noting the following: clisp-announce do not mail here without a prior discussion on clisp-devel or clisp-list clisp-devel subscriber-only, you &must-e; post from a subscribed address, no exceptions (see above) clisp-list used to be the only open list (it is closed now), so it is filtered especially aggressively: no &mime; mail (no &html; formatting, no attachments - send the URL instead) the list address must be in CC or TO, not BCC. do not mention virgin, penis or viagra in the Subject line. If you do not like this policy, please volunteer to maintain the mailing lists - you will be required to go through all the held for moderator's approval mail and approve/discard as appropriate at least twice a day. How do I report bugs? Patiently! How do I help? Please read and submit your patch, together with a src/ChangeLog entry (see other entries there for inspiration), to clisp-devel (if it is small) or SourceForge patch tracker. See src/CodingStyle for the style one should follow. The patch should be created using &hg;. How do I debug &clisp;? Passing &with-debug; to the top-level &configure; script does not just replace -O with -g in the &gcc; invocation, it also disables some optimizations (registers, assembly, generational &gc;or) and enables some printf calls (mostly before error messages caused by system calls). This means that some bugs might be exhibited by the normal build but ¬-e; by the &with-debug; one (and vice versa). If that is the case, please &configure; normally but edit the Makefile and replace the &gcc; optimization options (-O et al) with -g before running &make;. &sh-prompt; ./configure &with-debug; --cbc build-g &sh-prompt; cd build-g &sh-prompt; gdb lisp.run ;; or lisp.exe on &win32; and &cygwin; (gdb) boot (gdb) run &sh-prompt; ./configure &with-debug; --with-module=foo --cbc build-g &sh-prompt; cd build-g &sh-prompt; gdb full/lisp.run ;; or lisp.exe on &win32; and &cygwin; (gdb) full (gdb) run When debugging a &base; module, use base instead of full and boot above. If your &clisp; was built with generational &gc;or, you will see plenty of spurious segfaults (that's how the generational &gc;or works). At the end of src/.gdbinit you will find the following suggested &gdb; commands: handle SIGSEGV noprint nostop break sigsegv_handler_failed which will stop execution on a dangerous segfault. You can find out whether this is needed by examining the output of the following command: &sh-prompt; &clisp-cmd; --version | grep GENERATIONAL_GC You are encouraged to make use of commands defined in src/.gdbinit, such as boot base full run_test run_mod_test run_all_tests run_ansi_tests run_ansi_tests_compiled xout zout zbacktrace and others; help &command-r; in &gdb; will tell you what &command-r; does. &clisp; makes heavy use of &cpp; macros (e.g., pushSTACK), which can complicate debugging in &gdb;. The macros are defined in src/lispbibl.d, but the maze of conditionals may be hard to navigate. If you are using &gcc;, you can do make lispbibl.h and then examine the definitions in the generated lispbibl.h. See also bottom of src/spvw_debug.d. When you encounter a regression error (i.e., something worked in a previous version of &clisp;, but does not work in the current version) which we cannot reproduce, we will often ask you to figure out which specific &hg; commit introduced it. The way to figure that out is to use hg bisect. Why is &clisp; using menorah as the logo? Whimsical If you must have some answer and you do not care whether it is correct or not, you may simply think that &cl; brings the Light to a programmer, and &clisp; is a vehicle that carries the Light. Accordingly, &clisp; enables you to see the truth, thus you can pronounce it as see-lisp. Alternatively, if you are a seasoned expert, you might pronounce it as sea-lisp. Historical &clisp; has been using the menorah for the logo since the project was first started in the late 1980-ies by Bruno Haible and Michael Stoll. This probably reflects the authors' affection toward the Jewish people, Judaism or the State of Israel (neither of the two original authors is Jewish by birth). You may ask the original authors for details yourself. Both of them are very busy though, so do not expect a prompt reply. Shouldn't the logo be changed now due to the current political developments in the Middle East? The &clisp; developers, both the original creators and the current maintainers, do not subscribe to the mainstream view that blames the Jews for everything from high oil prices and Islamic extremism to El Niño and global warming (or cooling, whatever the looming disaster du jour is). Moreover, today, when Jews are being pushed out of the American and European academic institutions with various obscene boycott and divestment campaigns, it is crucial for all of us to stand together against the resurgence of Nazism. For more information, please see: Committee for Accuracy in Middle East Reporting in America Information Regarding Israel's Security Middle East Media Research Institute YES to peace, NO to terror More links Aren't there other political issues of concern? Yes, there are! For example, in 1989 the communist government of the People's Republic of China murdered some 3000+ student human rights protesters at the Tienanmen square in Beijing, and people appear to have already forgotten this crime. A note to that effect was kept in the file src/timezone.lisp until 2002, when it was decided that is was not visible enough there and it was moved here. We also oppose software patents and support other liberal (i.e., pro-liberty) causes. Aren't you afraid of losing some users who are offended by the logo? Do you have in mind people like this one? Good riddance! Using software to promote a political agenda is unprofessional! Expressing their opinion is a perfectly natural thing for the authors, be it artistic preferences, political views or religious beliefs. The use of the menorah has its roots somewhere between these areas, and the authors are proud to display it. If you are unlucky enough to have lost the freedom to express your opinion, due to the constraints of a government, society, religion, or expectations of professional relationships, the Free World condoles with you. The authors of &clisp; are not operating under such constraints. If you are unhappy about their artistic preferences, political views or religious beliefs, you are free to ignore them. Many scientists have been doing art, politics and religion. René Descartes and Isaak Newton combined mathematics and Christianity. Albert Einstein helped the U.S. to counter the danger of an atomic bomb in the hands of the Nazis. Bram Moolenaar, the author of VIM, promotes charitable donations to Uganda. Why are floats printed in binary? Where is &defun;? Pass &opt-M; to the &rt; (&lisp-run; or lisp.exe). Or, better yet, use the driver clisp (or clisp.exe) instead of invoking the &rt; directly. Where is the IDEIntegrated Development Environment? &emacs;-based run-lisp &slime; &ilisp; The Common Lisp Cookbook non-&emacs;-based VisualCLisp Jabberwocky GClisp &phemlock; What are the command line arguments? See . How do I get out of the debugger? See . What &clisp; extensions are available? bundledQuite a few modules are included with &clisp;, pass --with-module=... to &configure; to build them and see . 3rd party See the incomplete list of Common Lisp software running in &clisp;. DIYDo It Yourself See and for information on how to interface with external &c-lang; libraries. &http; &clisp; comes with src/inspect.lisp which implements a rudimentary &http; server CLOCC/src/donc/ has an &http; server CLOCC/src/cllib/ handles URLs mod_lisp hooks lisp into &apache; Object-Oriented HTTP server for &clisp; Both AllegroServe and CL-HTTP require multithreading and do not work with &clisp; yet. Where is the init (RC) file on my platform? Read the file <clisp.html#opt-norc> in your build directory or binary distribution (this is the version of the user manual for your platform). Where are the modules with which I built &clisp;? In the &full; &linkset;. Run &clisp; like this: &sh-prompt; &clisp-cmd; &opt-K; full Unless your &clisp; was configured with option &without-dynmod;, you should instead use (&require-my; &name-r;) where &name-r; is a &string-t;, e.g., &rawsock-mod;, or clx (for either &new-clx; or &mit-clx;, whichever is installed), or linux (for &glibc-mod;) etc. This is more reliable because the &full; &linkset; may include many modules which you do not need and, in fact, &cannot-e; use because they require external libraries not installed on your system. Making &base; the default &linkset; has some advantages: Shared Library Hell Avoid problems when a module requires a shared library, and that shared library is not present (or is present but with wrong version) on your system. Smaller Images Are Faster Adding things to the heap increases working set size causing a slowdown, which is especially important for small to medium applications. Uniform User Experience Composition of the &full; &linkset; is up to the packager, while &base; is the same on all platforms. See clisp-list for more information (/). How do I create a GUIGraphical User Interface for my &clisp; program? Use module >k2-mod;: create ui.glade with &glade;, then do &sh-prompt; ./configure --with-module=gtk2 --cbc build-gtk &sh-prompt; ./build-gtk/clisp &opt-K; full &opt-x; '(gtk:run-glade-file "ui.glade")' There are many other options, see "Common Lisp software running in &clisp;". How do I create an executable file with all my code in it? Use &savemem;, see also . When I deliver my application with &clisp;, does it have to be covered by the &gnu; &gpl;? Not necessarily. &clisp; is Free Software, covered by the &gnu; &gpl;, with special terms governing the distribution of applications that run in &clisp;. The precise terms can be found in the COPYRIGHT file contained in the source and binary distributions of &clisp;. Here is an informal clarification of what these terms mean in practice. Please refer to the said COPYRIGHT file when in doubt. In many cases, &clisp; does not force an application to be covered by the &gnu; &gpl;. Nevertheless, we encourage you to release your software under an open source license. The benefits of such a license for your users are numerous, in particular they are free to modify the application when their needs/requirements change, and they are free to recompile the application when they upgrade their machine or operating system. &clisp; extensions, i.e. programs which need to access non-portable &clisp; internal symbols (in the packages &sys-pac;, &clos-pac;, &ffi-pac;, etc), must be covered by &gnu; &gpl; as well. Other programs running in &clisp; have to or need not to be placed under &gnu; &gpl;, depending on their distribution form: Programs distributed as Lisp source or &fasl-file; files can be distributed without restrictions coming from &clisp;. Programs distributed as &clisp; &mem-image;s can be distributed only if accompanied with the non-&clisp; &fasl-file; files which make up the &mem-image;, and a &make-file; (or clear textual instructions) for rebuilding the &mem-image;. If you need to distribute a modified &clisp; executable (for example, incorporating additional &module;s written in &c-lang;), you must distribute its full source under &gnu; &gpl;. If you are not satisfied with this, you can instead put the additional &module;s into a separate (non-&clisp;) program, with which your Lisp program will communicate via &socket-stream;s. Where is the binary distribution for my platform? The &clisp; maintainers do not offer &clisp; binary distributions, we believe that software packagers specializing on a particular platform are in a better position to provide quality binaries. Indeed, the section Get CLISP on the &clisp;'s home page points to &clisp; packages for all major &linux; and &bsd; variants. The only exception is &win32; for which there is no centralized packager: usually, after each &clisp; release, a &clisp; user volunteers to build a &win32; binary package, which we then distribute at . But a previous release had a binary distribution for my platform, why does not the current one? It was probably contributed by a user who did not (yet?) contribute a binary distribution for the current release. You can find out who contributed a specific binary distribution by looking at the release notes in the &sf; Files section. Why does not &clisp; build on my platform? Please see file unix/PLATFORMS in your source distribution for information about some known troublesome platforms as well as instructions on porting &clisp; to new platforms. Why is the --with-libsigsegv-prefix option ignored by the &configure; script on my 64-bit machine? See a note on bi-arch systems in section 2 of unix/INSTALL in your source distribution (for the impatient: pass --libdir=='${exec_prefix}/lib64' to the &libsigsegv; &configure;). See these threads which discuss the situation in depth: What do these error messages mean: invalid byte #x94 in CHARSET:ASCII conversion and character #\u00B3 cannot be represented in the character set CHARSET:ASCII? This means that you are trying to read (invalid byte) or write (character cannot be represented) a non-&ascii; character from (or to) a character stream which has &ascii; &extfmt;. The default is described in . This may also be caused by filesystem access. If you have files with names incompatible with your &path-enc;, filesystem access (e.g., &directory;) will &signal; this &error-t;. You will need to set &path-enc; or pass &opt-E; to &clisp;. Using a &enc1-1;, such as &iso-8859-1;, should help you avoid this error. Note that this error may be signaled by the Print part of the &repl; and ¬-e; by the function you call. E.g., if file "foo" contains non-&ascii; characters, you will see such an error when you type (&with-open-file; (s "foo" :direction :input &extfmt; &iso-8859-1;) (&read-line; s)) If instead you type (&with-open-file; (s "foo" :direction :input &extfmt; &iso-8859-1;) (&setq; l (&read-line; s)) &nil;) &clisp; will just print &nil; and signal the error when you type l. What does this message mean: Display all 1259 possibilities? (y or n) &clisp; uses &readline; for command line editing and completion. You get this Display all 1259 possibilities message (and sometimes many screens of symbols) when you hit &tab-c; too many times in an inappropriate place. You can turn this feature off if you are using &emacs;. It is a good idea not to use &tab-c;s in your code. Why does not command line editing work? See . How do I avoid stack overflow? &clisp; has two stacks, the program stack and the lisp stack, and both may occasionally overflow. Generic You will always get a stack overflow when you try to print a circular object (&list-t;, &structure-object-t;, &vector-t; etc) and &print-circle-var; is &nil;. Just set &print-circle-var; to &t;. You will always get a stack overflow on infinite recursion. Some simple functions (like Ackermann's) recurse more than you might expect and can run out of stack on relatively small inputs. Compiled code uses less stack (and memory) and is faster. If you really do need more Lisp stack, you can increase it by telling &clisp; to pre-allocate more memory using &opt-m; or &make-thread;. If you get a segmentation fault after (or instead of) a program stack overflow message, please make sure that you had &libsigsegv; installed when you configured and built &clisp;. Specific &win32-only; You have several option, use whichever is applicable to your situation:modify the global SYSTEM.INI modify the PIF that you use to invoke &clisp; (/) or set the program stack (of the run times base/lisp.exe or full/lisp.exe, ¬-e; the driver clisp.exe) using editbin (/); you can find the current stack size using dumpbin. &unix-only; Increase program stack with ulimit -s (or limit stacksize if you are using csh) to at least 8 MB. Why does my function return different values on each invocation and then hangs forever? The following code modifies itself: (defun my-func ()modifies internal data! (let ((var '(a b c))) (nconc var '(1 2 3)))) and will not work as one would naively expect: (&setq; &print-circle-var; &t;) (&function-lambda-expression; #'my-func) (LAMBDA NIL (DECLARE (SYSTEM::IN-DEFUN MY-FUNC)) (BLOCK MY-FUNC (LET ((VAR '(A B C))) (NCONC VAR '(1 2 3))))) (my-func)the first invocation (a b c 1 2 3)modifies the internal data: (&function-lambda-expression; #'my-func) (LAMBDA NIL (DECLARE (SYSTEM::IN-DEFUN MY-FUNC)) (BLOCK MY-FUNC (LET ((VAR '(A B C . #1=(1 2 3)))) (NCONC VAR '#1#))))note shared structure! (my-func) (A B C . #1=(1 2 3 . #1#))circular list! (function-lambda-expression #'my-func ) (LAMBDA NIL (DECLARE (SYSTEM::IN-DEFUN MY-FUNC)) (BLOCK MY-FUNC (LET ((VAR '(A B C . #1=(1 2 3 . #1#)))) (NCONC VAR '#1#))))note circular value of var (my-func)hangs because &nconc; cannot handle a circular argument Instead you must do (defun my-func () (let ((var (copy-list '(a b c)))) (nconc var (copy-list '(1 2 3))))) &disassemble; will show the constants in your compiled functions (see for the explanation of the &disassemble; output.) As you see above, &function-lambda-expression; works for the interpreted ones. See Lisp Programming Style for more useful information. Why don't floating point arithmetic return what I want? (- 1.1 0.9) 0.20000005 inexactFloating point arithmetic is inherently inexact, so this ¬-e; a bug, at least not a bug in &clisp;. You may argue that this is a bug in &ieee-bfp;, but, before you do, please make sure that you do know What Every Computer Scientist Should Know About Floating-Point Arithmetic (or are at least familiar with the light version thereof). See also clisp-list (/). exactIf you want exact calculations, use &rational-t;s: (- 11/10 9/10) 1/5 Why does &sh-prompt; &clisp-cmd; &opt-x; '(&random; 1s0)' always print the same number? Reproducibility is important. See . Why is an extra line break inserted by the pretty printer? See . For example, here you are pretty-printing two objects: a one-liner #\Q and a two-liner &cr-c;+&lf-c;, and a line break is inserted between them so that they are printed on separate lines: (defparameter crlf (coerce '(#\Return #\Linefeed) 'string)) CRLF (map 'list #'char-code (format nil "Q~a" crlf)) (81 10 13 10)not (81 13 10) ! When you want &cr-c;+&lf-c;-terminated lines, use &extfmt; &dos-k; argument when opening your &stream-t;s. How do I disable this annoying warning? &clisp; often issues &warning-t;s when it encounters suspicious behavior. It is best to fix your code rather than to suppress them. To figure out where the warning is coming from, do (&setq; &break-on-signals-var; '&warning-t;) and examine the stack in the &debugger; to see where the warning is coming from. If everything else fails, read the manual. Why does &defvar; affect previously defined lexical closures? (defun adder (val) (lambda (x) (+ x val))) ADDER (setq add-10 (adder 10)) ADD-10 (funcall add-10 32) 42as expected (defvar val 12) VAL (funcall add-10 0) 12why not 10?! Explanation The above code does not conform to &ansi-cl;, therefore &clisp; can produce arbitrary results. See . Remedy Always follow the naming convention for global special variables defined by &defvar; and &defparameter; (e.g., *FOO*) and &defconstant; (e.g., +BAR+). More / / The error message is not helpful! Sometimes an error message contains a compound object whose content you want to examine. Often this object will be available for &inspect-my;ion via the Inspect command (see ). E.g., *** - READ: input stream #<INPUT STRING-INPUT-STREAM> ends within an object The following restarts are available: ABORT :R1 ABORT Break 4 [1]> :i #<SYSTEM::SIMPLE-END-OF-FILE #x000333CB82F8>: standard object type: SYSTEM::SIMPLE-END-OF-FILE 0 [$STREAM]: #<INPUT STRING-INPUT-STREAM> 1 [$FORMAT-CONTROL]: "~S: input stream ~S ends within an object " 2 [$FORMAT-ARGUMENTS]: (READ #<INPUT STRING-INPUT-STREAM>) INSPECT-- type :h for help; :q to return to the REPL ---> 0 #<INPUT STRING-INPUT-STREAM>: atom type: STRING-STREAM class: #1=#<BUILT-IN-CLASS STRING-STREAM> INSPECT-- type :h for help; :q to return to the REPL ---> :d #<INPUT STRING-INPUT-STREAM> is an input-stream. It reads from "(foo" from 0 to 4 at 4. Why is the function FOO broken?! When confronted with unexpected behavior, try looking in the . E.g., &clisp; &directory; is different from the &cmucl; implementation, so, to find out how to get the results you want, you should search the &clisp; implementation notes for "DIRECTORY". Alternatively, since the implementation notes are organized in parallel to the &ansi-cl;, and &directory; belongs to the Chapter 20 in &ansi-cl;, you can just go to in impnotes and look for "DIRECTORY" there.