gnuplot-beta

New bug: "save" trashes "load" behavior

[Ethan M. <merritt@u.washington.edu>]

I can't figure this one out:

Terminal type set to 'x11'
gnuplot> 
gnuplot> 
gnuplot> show loadpath

        loadpath is 
        system loadpath is "../demo" 

gnuplot> load 'charset.dem'
gnuplot> save 'xxx'
gnuplot> load 'charset.dem'
                           ^
         Cannot open load file 'charset.dem'
         util.c: No such file or directory

gnuplot> show loadpath

        loadpath is 
        system loadpath is "../demo" 


After doing a "save", I can no longer do a "load" from the
loadpath directory.  Loads from the current directory still work.

Any ideas?  


-- 
Ethan A Merritt       merritt@u.washington.edu
Biomolecular Structure Center
Mailstop 357742
University of Washington, Seattle, WA 98195

Re: New bug: "save" trashes "load" behavior

[Ethan M. <merritt@u.washington.edu>]

On Tuesday 15 February 2005 12:37 pm, Ethan Merritt wrote:
> 
> After doing a "save", I can no longer do a "load" from the
> loadpath directory.  Loads from the current directory still work.

Tracked it down partially, but I don't know the proper fix:

in variable.c line 157, the loadpath_handler sets
	limit = envptr

After that, the loadpath code will never again look at
the part of the loadpath that came from the environment.

This is not a part of the code I have ever looked at
before, so I am uncertain what the intended flow of
control is.

-- 
Ethan A Merritt       merritt@u.washington.edu
Biomolecular Structure Center
Mailstop 357742
University of Washington, Seattle, WA 98195

Re: New bug: "save" trashes "load" behavior

[Hans-Bernhard B. <br...@ph...>]

Ethan Merritt wrote:
> On Tuesday 15 February 2005 12:37 pm, Ethan Merritt wrote:
> 
>> After doing a "save", I can no longer do a "load" from the loadpath
>>  directory.  Loads from the current directory still work.

You called this a "new bug" in the subject --- that turns out to have
been a bit misleading.  I didn't actually run back-version tests, but as
far as I can see, this is a genuine bug in loadpath_handler(), and has
been there ever since the module was first written by Lars(?).

The theory, as far as I understand it, is this:  'save' wants only the
'set loadpath' part of the path, so the case ACTION_SAVE: entry puts
'limit' to that position in the full path where the $GNUPLOT_LIB path
begins, and thus the fall-through ACTION_GET will stop there.  The bug
is that on reaching the limit, it's not reset, so it'll stay in place
forever.  The fix is to re-initialize 'limit' to NULL at the same time p
is initialized to NULL.

I'm checking in the fix right away.