- 30 Jan, 2021 40 commits
-
-
Wolfgang Dautermann authored
Most default values are rendered as @code{...}. Improvement of commit [e4199e].
-
Wolfgang Dautermann authored
-
Gunter Königsmann authored
-
Leo Butler authored
bug #3154: lratsubst NOT as described in Help file Valid first arguments to lratsubst are now correctly described in the texinfo documentation and the code has been modified to conform with the behaviour of subst. bug #3376: fullratsubst doesn't catch infinite recursions fullratsubst1 (where the infinite recursion happened) has been rewritten using a loop. See also below. bug #3706: lratsusbst causes bind stack overflow on large lists lratsusbst1 has been rewritten to loop over the argument list to prevent; it also loops over the substitutions if fullratsubstflag=true. An upper limit on the number of iterations, lrats_max_iter, is used to prevent infinite loops. The documentation for lratsusbst and fullratsubst has been updated to demonstrate the new behaviour; lrats_max_iter is also documented. 4 tests have been added to tests/rtest6a.mac. --- I would like to say that all testsuite tests pass, but... there were existing testsuite errors in rtest_descriptive.mac and rtest_abs_integrate.mac that are still there (gcl, ecl).
-
Barton Willis authored
o Append regression tests All tests pass using Clozure CL
-
Leo Butler authored
This is the version from commit 7a920ade9 from the texinfo repository. It includes a patch that is functionally equivalent to that recently introduced here. This overwrites the patch in fea7e4f87. category-macros.texi has a patched version of \mmref that is functionally equivalent to its predecessor. This amends 942fc018e. Ref: https://lists.gnu.org/archive/html/bug-texinfo/2021-01/msg00011.html
-
Wolfgang Dautermann authored
-
Leo Butler authored
Commits a6b0663d4..942fc018e have removed the need for the special-casing.
-
Leo Butler authored
-
Leo Butler authored
In commits e1527087 and 57819f2f, Jinsong Zhao's patch to texinfo.tex's \ref macro was implemented. Wolfgang Dautermann noticed the pdf manual had "funny" characters appearing in place of _ (and others). This was caused by Zhao's patch--it was not properly isolating its catcode changing. Rather than maintain a separate macro for \ref, I have patched \ref in texinfo.tex (and asked the patch be adopted upstream) so that it can be customized to do what the patch in the above-mentioned commits should have done. Some additional changes: -\mmref was defined as a 2-arg macro, but the second argument was never used, so it is dropped; -\mmref needs to be defined before its dependents, so the tex block at the end of the file is moved above the definition of @mref; -the current patch removes the need for @mrefdot & co.; they have been left in, but redefined to use the correct parent (@mref) rather than duplicating code.
-
Leo Butler authored
In bug report #3549, Jinsong Zhao's patch obviated the need for these files, but they were retained. They complicate the build process for maxima.pdf and are no longer needed with a patched version of mmref.
-
Leo Butler authored
With this switch, there is no need for the (faulty) overwriting of the texinfo macro \setpdfdestname in category-macros.texi nor for mmref.tex. This has been proposed as a patch to the texinfo maintainers: https://lists.gnu.org/archive/html/bug-texinfo/2021-01/msg00010.html
-
Leo Butler authored
The patch to \var is not needed with the patched \mmref macro.
-
Kris Katterjohn authored
* Move 2 entries from share to core where they belong * Consolidate the 3 entries about improved compatibility * Move the generic entries to the end of the section * Add 2 numbered bugs related to the translator * Sort the numbered bugs by number * Capitalize some words These are just some things I noticed while skimming the change log. I didn't attempt to reword or clarify existing entries.
-
Kris Katterjohn authored
An error could occur in FREE-LISP-VARS due to a backquoted form from the translation of MFACTORIAL. Whether or not an error occurred was implementation dependent because it depended on how the backquote reader macro is implemented. I observed the error with SBCL, but not with ECL or CLISP. So eliminate that backquoted form from the translation of MFACTORIAL so the naive FREE-LISP-VARS doesn't choke on it. No problems with the test suite or share test suite. tests/rtest_translator.mac runs as expected, with new tests.
-
Wolfgang Dautermann authored
@var{|x|} was printed as "— x —" in the PDF manual.
-
Leo Butler authored
Add bugfixes for #3700 and #3674.
-
Leo Butler authored
ssubst was written recursively, which can lead to stack overflows if the compiler does not replace the tail recursion with a loop. Re-write the function using a while loop to fix this. A test is added to rtest_stringproc.mac.
-
Leo Butler authored
Commit a8d0f5db72 introduced the latex property of the output text, which contains the latex code used to generate the overlayed image. The current commit updates imaxima-latex to use that stored latex code. This simplifies export of the *imaxima* buffer.
-
Wolfgang Dautermann authored
See: https://sourceforge.net/p/maxima/mailman/message/37181353/ and: https://sourceforge.net/p/maxima/mailman/message/37181827/ for further information about the issue and the solution. I patched the updated version from the GNU website, Leo's patch worked properly, after changing the version number to: \def\texinfoversion{2019-09-28.10}. The manuals look ok. Especially the PDF manual looks better now, thanks Leo!
-
https://www.gnu.org/Wolfgang Dautermann authored
The patch from Leo Butler to fix some PDF problems (next commit) can be applied and it works too, so let's update first.
-
Wolfgang Dautermann authored
-
Wolfgang Dautermann authored
(using admin/list_authors.pl, which evaluates all Git commits).
-
Gunter Königsmann authored
Documents commit df47fb385d891
-
Robert Dodier authored
and call rat or fullratsimp on all arguments. Fixes SF bug #3658: "error with floats appearing instead of integers in lratsubst"
-
Robert Dodier authored
Fixes SF bug #3434: "kill(ratvars) should give an error"
-
Wolfgang Dautermann authored
Mozilla and Netscape are no longer really relevant, Google Chrome is.
-
Wolfgang Dautermann authored
This commit mentioned only the Windows installer related changes, and is superseded with the Changelog Gunter created with committed [06f0cb]. Kept the missing changelog entries in 'make dist'. And used my description for the Windows related changes, for an ordinary user the NSIS syntax 'RequestExecutionLevel admin' is not really important.
-
Wolfgang Dautermann authored
-
Gunter Königsmann authored
This is open source so anyone who can improve it is invited to do so.
-
Barton Willis authored
o New tests o insert some intl:gettext stuff
-
Kris Katterjohn authored
The commit message for commit d7ff7c39 claims it was only inlining a function, but that commit made several other changes as well. One of the changes was using = instead of EQUAL in GCDISP. The use of = here causes various things like gcfactor(x*%i) and gcfactor(%i/2) to signal lisp errors. I'm changing this = to EQL in GCDISP. (EQUAL isn't necessary here; the original use of EQUAL goes back to the Maclisp days when EQL didn't exist.) No problems with the test suite or share test suite. New tests are added to rtest_numth.
-
Gunter Königsmann authored
Main advantage of this update: Nicer display of sets.
-
Gunter Königsmann authored
This resolves a small privacy issue: If the user still used the old Internet Explorer 9 as help browser and opened the wxMaxima manual the manual downloaded a HTML5 compatibility shim over the internet. Debian deems downloading a file without asking the user first as a privacy issue
-
Wolfgang Dautermann authored
-
Andreas Eder (are_muc) authored
be more efficient.
-
Kris Katterjohn authored
This is an ancient bug that was present in Macsyma. When translating a conditional expression with a test that translated to T and consequent that translated to NIL, that entire clause was being skipped over. An example of the bad behavior: the translation of if true then false else 'wtf would yield wtf instead of false. It was not necessary for the test and consequent to be a literal true and false, respectively; these just needed to translate to T and NIL. Since this was around in the old Macsyma days, my guess is that this was a careless attempt to save some conses. The idea being that we can omit the final clause in (COND ... (T NIL)) The problem is that the code to handle this optimization was looking at all of the clauses instead of just the final one. We'll keep this tiny optimization for now, but we'll restrict this to the final clause. No problems with the test suite or share test suite. tests/rtest_translator.mac runs as expected, with new tests.
-
Kris Katterjohn authored
This is an old bug that was present before Maxima 5.0. It doesn't appear that this bug was present in Macsyma because I haven't found any evidence that elseif existed in Macsyma. The translation of a conditional with another conditional directly nested in an elseif clause has been totally wrong. Using else-if (else if) instead of elseif would work fine. An example of some particularly bad behavior: if false then 'wtf1 elseif false then if true then 'wtf2 else 'wtf3 else correct was being incorrectly translated like if false then 'wtf1 else if true then 'wtf2 else 'wtf3 So this would yield wtf2 instead of correct. Some portions of code were not updated to correctly handle elseif when it was first introduced. See commits d89c9965 and 4cc510ca for some late fixes related to elseif. The problem in the translator has been due to an optimization for the else-if case. The check for when to apply the optimization was not particularly careful and the introduction of elseif caused the translator to incorrectly apply the optimization technique to elseif clauses like the one above. The optimization is still useful for the else-if case, so we're keeping it. We're just more careful about when to apply it now. No problems with the test suite or share test suite. tests/rtest_translator.mac runs as expected, with new tests.
-
Kris Katterjohn authored
Some upcoming commits will change how MCOND is translated
-
Kris Katterjohn authored
The false vs NIL behavior this comment referred to was changed in commit e43c7041 in 2005.
-