problem with interface between evolver and geomview

[sullivan at msri.org (John Sullivan)]

I have occasionally encountered a problem where geomview doesn't delete
objects , but instead just accumulates many copies of them, as the evolver
evolves the data.  I finally realized that this is happening exactly when
I give the evolver a non-trivial path for my file name.

If I invoke evolver on, say "./foo", then it sends the command
"(geometry ./foo ...)" to geomview, and then later, if I switch
file names (by saying "q", and then perhaps "foo" or "././foo"
to the evolver) it tells geomview "(delete ./foo)", which geomview
seems to complain about:
delete: id expected in arg position 1 (got "./foo")
(Note that my TARGET browser in geomview shows simply "foo" as the
name of the object.)

I can't quite figure out what's going on.  When I use "P\nA" in
the evolver, and then "cat /tmp/foo.a##### | more" to see what it
sends, it always looks like it sends the full path name I've used.

I just tried "evolver ././foo", followed by "P\n8" to invoke geomview.
I got an object foo.  Then "q\n./foo" to evolver" made me end up
with an object that geomview calls "o".  (It properly deleted foo.)
Then "q\nfoo" seems to put the new geometry under the same name "o".
Then "q\n././foo" deletes "o" and leaves me with a TARGET "foo".
I had a hard time reproducing the case where I get the "id expected"
message this time, but here's one full evolver session that did it.
(There are much shorter ones, I'm sure, but I didn't happen upon it
this time.)

    riga[114]  evolver ././foo
    Surface Evolver Version 1.92j December 11, 1993

    Enter command: P
    Choose output graphics format: 
    1. Pixar file 
    2. MinneView (OOGL) file
    3. PostScript file
    4. Triangle file 
    5. Softimage file 
    8. Start simultaneous geomview  
    9. End simultaneous geomview  
    A. Start OOGL pipe (you must start reader)  
    B. End OOGL pipe
    0. cancel  
    Choice: 8
    Enter command: q
    Enter new datafile name (none to continue, q to quit): ./foo
    Enter command: q
    Enter new datafile name (none to continue, q to quit): foo
    Enter command: q
    Enter new datafile name (none to continue, q to quit): ././foo
    Enter command: q
    Enter new datafile name (none to continue, q to quit): foo
    Enter command: q
    Enter new datafile name (none to continue, q to quit): ./foo
    Enter command: q
    Enter new datafile name (none to continue, q to quit): ././foo
    Enter command: q
    Enter new datafile name (none to continue, q to quit): ./foo
    Enter command: q
    Enter new datafile name (none to continue, q to quit): foo
    Enter command: q
    Enter new datafile name (none to continue, q to quit): ./././foo
    Enter command: q
    Enter new datafile name (none to continue, q to quit): ./foo
    Enter command: delete: id expected in arg position 1 (got "./foo")
At this point I'm left with 
[g1] foo
[g2] o
in my TARGET browser.

I'm tempted to classify all of this as geomview bugs, not evolver bugs,
though all you experts may care to disagree.  As a work-around,
I'd suggest to Ken to strip off leading path components when forming an
object name for geomview.

-John