- 16 Nov, 2019 1 commit
-
-
Robert Dodier authored
Fixes bug reported to mailing list 2019-11-14: "get('absimp,'version) always returns false."
-
- 14 Nov, 2019 2 commits
-
-
-
Kris Katterjohn authored
In Maclisp and Lisp Machine Lisp there was a variable named PRIN1 which, if non-NIL, specified a function for the toplevel REPL to call instead of PRIN1. This variable did not affect the output from other functions that would call PRIN1 (like PRINT). When non-NIL, the value of PRIN1 was explicitly called by Macsyma for just one specific thing: printing trace output when the lisp_print trace option was set. Some time before Maxima 5.0 a PRIN1 variable was defined in commac. Earmuffs were put on in commit ad12e542 in 2004. Since the value of *PRIN1* has not been used by the underlying lisp and its value has still only been used in this one very specific place, I'm just removing it altogether. No problems with the test suite or share test suite.
-
- 13 Nov, 2019 1 commit
-
-
Kris Katterjohn authored
ambigious -> ambiguous
-
- 11 Nov, 2019 1 commit
-
-
Wolfgang Dautermann authored
-
- 09 Nov, 2019 3 commits
-
-
Kris Katterjohn authored
The final clause in a COND form was unreachable because the previous clauses covered the cases of FORM being either an atom or a cons. In PDP-10 Maclisp and Franz Lisp that corresponding clause was reachable because their hunks are neither atoms nor conses. No functional changes.
-
Kris Katterjohn authored
After commit f6fb5fb9 PUSH-AUTOLOAD-DEF is unused. See that commit message for more about this function. No problems with the test suite or share test suite.
-
Kris Katterjohn authored
If FOO had the AUTOLOAD property then something like (PUSH-AUTOLOAD-DEF 'FOO '(BAR BAZ ...)) would construct a form to give BAR, BAZ, etc. the same AUTOLOAD property and it would push this form onto *PRE-TRANSL-FORMS*. (If FOO did not have the AUTOLOAD property anymore at the time the constructed form got evaluated, then the AUTOLOAD property on BAR, BAZ, etc. was just T instead.) The calls we've actually had to PUSH-AUTOLOAD-DEF haven't done anything useful because none of these symbols have the AUTOLOAD property anymore: MARRAYREF, MARRAYSET, $APPLY1, $APPLYB1, $APPLYB2, $CF and $SCANMAP. All of these symbols did have the AUTOLOAD property at one time (on at least PDP-10 Macsyma). (Side note: since PUSH-AUTOLOAD-DEF was called for $APPLY1, $APPLYB2 and $APPLYB2, I find it strange that it was not called for $APPLY2.) Since none of those symbols have had the AUTOLOAD property, the calls to PUSH-AUTOLOAD-DEF would just return NIL without doing anything else. No functional changes.
-
- 08 Nov, 2019 1 commit
-
-
Kris Katterjohn authored
apply2 and applyb2 used to use a special variable *RULELIST to pass data to APPLY2 and APPLY2HACK, respectively. Commit 23d46f5a from 2013 attempted to eliminate this special variable by passing an extra argument to APPLY2 and APPLY2HACK instead; however, the translations of apply2 and applyb2 were never updated. A translated apply2 would still bind *RULELIST and then call APPLY2 with the wrong number of arguments. A translated applyb2 would also still bind *RULELIST and then call APPLY2HACK with the wrong number of arguments -- but that's not all. APPLY2HACK returns a cons whose car is the value to return from applyb2 and whose cdr is an integer (the value APPLY2HACK compares to maxapplyheight). A translated applyb2 would actually just return the cons itself, and various parts of Maxima (like SIMPLIFYA) would choke on this. Besides trivial downcasing and whitespace changes, the translations of apply2 and applyb2 haven't changed in decades. So while the translation of apply2 has only been broken since 2013, it appears that the translation of applyb2 has always been broken. I can't find any evidence that applyb2 was ever translated correctly before now.
-
- 06 Nov, 2019 1 commit
-
-
Kris Katterjohn authored
After commit 270ba62c in 2009, which fixed bug #464, there was a superfluous test before signalling an error. The (NOT (EQ X Y)) test was always true at that point because (EQ X Y) is the test form for the previous COND clause. I noticed this, but Stavros Macrakis actually mentioned this in his initial report of bug #464 in 2003. No functional changes.
-
- 02 Nov, 2019 3 commits
-
-
Kris Katterjohn authored
natural_number was listed twice and complex was not listed at all. Also fix a whitespace issue.
-
Kris Katterjohn authored
See also commit 6809a3b9. No problems with the test suite or share test suite.
-
Kris Katterjohn authored
Discussed on the mailing list circa 2019-05-26. No problems with the test suite or share test suite.
-
- 01 Nov, 2019 2 commits
-
-
Kris Katterjohn authored
This is consistent with how factor and gfactor mark their results with the FACTORED and GFACTORED flags, respectively. No problems with the test suite or share test suite.
-
Kris Katterjohn authored
As Stavros Macrakis suggests in the bug report, we now mark the result as simplified so things like gcfactor(9) yield 3^2 instead of 9. No problems with the test suite or share test suite.
-
- 29 Oct, 2019 2 commits
-
-
Kris Katterjohn authored
If A and B are both 0 then their gcd as computed by GCD is 0. We need to check for this before dividing A and B by their gcd. gcfactor(0) was handled correctly in Macsyma, but it looks like this behavior changed at some point before Maxima 5.0. No problems with the test suite or share test suite.
-
Kris Katterjohn authored
-
- 28 Oct, 2019 2 commits
-
-
Wolfgang Dautermann authored
-
Robert Dodier authored
-
- 27 Oct, 2019 8 commits
-
-
Kris Katterjohn authored
See also commit 50558823 for more on zunderflow.
-
Kris Katterjohn authored
ZUNDERFLOW was a Maclisp variable that specified whether or not floating point underflows would be treated as errors. zunderflow has been an undocumented alias for ZUNDERFLOW since the documentation for zunderflow was removed in commit a07aedae in 2004.
-
Kris Katterjohn authored
nopoint has been an undocumented alias for *NOPOINT. See commit 2cfa6129 for more on *NOPOINT.
-
Kris Katterjohn authored
*NOPOINT was a Maclisp variable that specified whether or not to print a trailing decimal point after integers when the output base was 10: in this case *NOPOINT having the value T was like *PRINT-RADIX* having the value NIL in Common Lisp.
-
Kris Katterjohn authored
A constant literal list is iterated over when setting up some aliases and reverse aliases, but some of this constant data has been unused since commit 171b42bb in 2008. No functional changes.
-
Kris Katterjohn authored
-
-
Robert Dodier authored
* add @deffnx modedeclare so alias can be found via ? and ?? * cut out mention of string, even, and odd in list of modes since those are not actually recognized The documentation for mode_declare needs a lot of work so this is a small step in that direction.
-
- 26 Oct, 2019 6 commits
-
-
Kris Katterjohn authored
The BAK-TOP-FORM macro seemed to first appear in Maxima 5.4beta1 (circa 1998). It's just an identity macro and that's only used a couple of times in src/mdebug.lisp. No functional changes. No problems with the test suite or share test suite.
-
Kris Katterjohn authored
BAKTRCL (now *BAKTRCL*) was originally a (lisp) list that kept some backtrace information in Macsyma. Its default value was NIL and it was bound in MEVAL* and MEVAL1. When DEBUG (now *MDEBUG*) was set to all, MEVAL1 would push certain forms it was evaluating onto BAKTRCL. In ERRBREAK, which was an ERRSET handler, a variable named backtrace was set to a (Maxima) list containing the elements of BAKTRCL. This remained the same through Maxima 5.2 (circa 1995). In Maxima 5.4beta1 (circa 1998) src/mdebug.lisp appeared. A function named bt was present which could be used to print some backtrace information held in BAKTRCL, although it seemed to be a work in progress. Either way, the code in MEVAL1 to push onto BAKTRCL was also commented out and so BAKTRCL was always NIL anyway. The commented-out code in MEVAL1 was removed in commit 66a62cf7 in 2007 and BAKTRCL was renamed *BAKTRCL* in commit f4931e42 in 2009. Since then, *BAKTRCL* still gets bound in MEVAL* and MEVAL1 but it is always NIL. The function bt is undocumented and unused in src and share. It always just returns NIL since *BAKTRCL* is always NIL. No problems with the test suite or share test suite.
-
-
Robert Dodier authored
Fixes SF bug report #3521: "simplify_sum(sum(binomial(1,n),n,0,inf)) gives 3 instead of 2"
-
Kris Katterjohn authored
This function just calls GET with its arguments. There used to be a read-time conditional to implement this function differently on Lisp Machines. The conditional was removed in commit d72823ee in 2007. No functional changes.
-
Kris Katterjohn authored
The comment stated that the code was written the way it was so that TRANSLATOR-EVAL and TRANSLATE-MACEXPR-TOPLEVEL could still be traced when PURE was set to NIL. PURE was a (PDP-10 only) Maclisp variable which was related to read-only memory and UUO links. See commit 47dcc8a0 for a little more on UUO links and how they're related to tracing. Also fix an incorrect comment in this function. No functional changes.
-
- 24 Oct, 2019 1 commit
-
-
Wolfgang Dautermann authored
wxMaxima now requires more software for building it.
-
- 23 Oct, 2019 1 commit
-
-
Robert Dodier authored
-
- 22 Oct, 2019 2 commits
-
-
Robert Dodier authored
Conflicts: tests/rtest3.mac tests/rtest_tex.mac Resolve conflicts by accepting changes from both sides.
-
Robert Dodier authored
Reapply commit d93681aa (i.e. reverting revert 11ee6d57) for work in progress on display index properties.
-
- 20 Oct, 2019 3 commits
-
-
Robert Dodier authored
(2) Make declare_index_properties an argument-quoting function.
-
Robert Dodier authored
-
Robert Dodier authored
-