CVS log for loncom/lonmaxima

Up to [LON-CAPA] / loncom

Request diff between arbitrary revisions

Keyword substitution: kv
Default branch: MAIN

- s-nail requires space instead of comma as separator between email addresses.

- Replace use of mailto (metamail package) with mail command.
  metamail was in Red Hat 7.3, but is absent from Fedora, CentOS, RHEL etc.

- Replace use of mailto (metamail package) with mail command.
  metamail was in Red Hat 7.3, but is absent from Fedora, CentOS, RHEL etc.

See BUG# 6653 for a detailed description.
Fixes sporadic occurrences of "code ran too long" whenever maxima was used
in problems.

In case of a syntax error maxima 5.23 outputs "incorrect syntax" instead of the uppercase "Incorrect syntax" from previous versions.

Deal with multiline output

reset() resets display2d to true which is not what we want

We are quite often using the simp flag in maxima. It is quite handy for problems of type "simplify the following expression". However, simp is not reset to its default value (true) by the kill(all) incantation. reset() does the job.

If reset is not wanted here for other reasons, I recommend replacing it by
simp:true;
in order to do the job intended here.

Bugfix:
The sync-call before the for loop is premature. For at this time the connection to the server has not been accepted. Hence, the sync-call catches a timeout-error message. Unfortunately this takes 20s. So the client will be set up with a delay of 20s. Calls to maxima in this timeslot might lead (and do in practice) to "code ran too long" errors in the calling problem.

Moving the said sync-call after the $server-accept command remedies this issue.

- cap how many extra children there can be in case of oddities

- BUG#5517 when a child is exiting let the parent now so it can fork off a new child right away without having to wait for the child to fully pass on

- BUG#5517 try to get maxima to exit more quickly by explicitly quit()ting

- sles + maxima 5.9.1 + clisp + expect 1.15-293.1 (ie s8)
     - initial sync up was not occuring, use the full sync mechanism to
       get past the satrup messages and find the first set of
       command prompts

- ingores all output up to the first output label, then grabs all output from there on out.

Critical bug removed:
For multiline Maxima "%o"-output, only the first line has been used. Now, all lines will be processed.

Example:
diff((sin(x)^2-cos(x^2))/sin(x),x,1);

lonmaxima.session.log:
(%o2) (2*x*sin(x^2)+2*cos(x)*sin(x))/sin(x)
       -cos(x)*(sin(x)^2-cos(x^2))/sin(x)^2

LON-CAPA:
(2*x*sin(x^2)+2*cos(x)*sin(x))/sin(x)
Only the first line has been passed over to LON-CAPA.

- in some cases maxima just goes off and spins and spins, soft_close
  never returns etc. so taking the forceful way at when exiting

reset in MAXIMA has sometimes unexpected/unpredictable results.
Version 1.22 had it taken out before.

- fuller reset
- set the display to not have it do the 2d ascii art

- explicitly look for the (%i1) prompt after the kill(all) reset of the env
  otherwise it waits forever looking for the wrong i number

If 'Incorrect syntax' is detected - stop computations.
Return 'Incorrect syntax' as reply.

Alternative solution to the case of slow computations.

while() loop eliminated in favor of use of a regular expression in $command->expect() which includes next input prompt. Output is only parsed for ouput (o%N) from previous executed statement (i%N), once next input prompt (i%N+1) detected.

Timeout included in $command->expect() call will now terminate computations which take too long.

One solution to slow processing of computations -make this a while() loop.

Tested on FC5 and FC6: 32 bit/64 bit using perl-Expect, perl-IO-Tty, gmp, maxima, maxima-runtime-gcl and sbcl from fedora extras.

^M (Windows-style) line breaks needed to be removed.

Results of computations were being reported by maxima after $reply to client had already been set (as ''). Caused by duplicate input line ($i\d+) entry: e.g. from maxima transactions:
(%i2) trigsimp(trigreduce(4*a*x^3-4*a*x^3));
(%i2) trigsimp(trigreduce(4*a*x^3-4*a*x^3));
(%o2)                                  0

Now check for output line and make two passes if necessary. (Should this be a while() loop for the case where N passes are needed?).

Substitution of \(%o\d+\) label now functional.

The reset()-command in MAXIMA is broken
Need to actually do string replacement

Open3 is not needed.

Did not work with old Maxima version.

Don't make the client wait for the synching, do it as cleanup instead.

Need to sync up in order to deal with unexpected output or error situations.

Error processing, resetting

YES!

The latest on running the server

At least process management now works, but Maxima behaving oddly.

Even more interesting, still defunct.

Does a lot of interesting things, except work

PID of Maxima process needs to be global, so timeout handler has it when
killing the process

- need to install the CHLD signal handler before we fork the children
  off so REAPER can handle the maxima children deaths

- check for ref status before trying to use them as such

- eliminating unneeded globals

Works.

- some more style

- globals cleanup
- some style

- some style police

- yes we can use strict

Playing some more, having fun

Playing around with Computer Algebra System