Re: [Echempp-devel] recent model version

[Kai L. <kai...@un...>]

Hello Bernd,

thanks for testing... I'll check all your comments
(quite a lot :) in the next days.
However, I couldn't reproduce some
of your problems (see below).
Therefore, due to the "technical" problems,
could you do a complete and fresh installation:
top level "make clean" + remove all epp-libraries from
your <prefix>/lib directory. Then run a top level
"make install". Please tell me if something has changed.

> (1) I can easily and quickly start model from my
> /home/bs/software/EChem++ directory, however, if I try to start from my
>home directory, the program uses up all resources of the notebook, and
even >after several minutes, I do not get anything more than an empty
frame. If I >try to start from /home/bs/software or
> /home/bs/some_other_dir it seems to be somehow intermediate.

I can run the model-exe from my <prefix>/bin directory
(after "make install" !!) without any problems.
Today, I have created a link to this program on my
desktop and it still starts quickly and works as expected.
(certainly, we need a complete installation docu for
 the frist release, but this is still in work)
However, the wrapper script in the GUI/Windows/Qt/EChem++/Model
directory seems to need all system resources at launch time.
So please, use the freshly installed executable in your
<prefix>/bin directory for all subsequent tests.
I will try to find something about the wrapper.

> (2) I can use
> xxx> model >& /dev/null &
> to start the program in the background and sending all message output t=
o
>the null device. This is very good, since it could be used with an
icon, >avoiding all the messages being printed to some window.

Ok - that is a proper possibility when calling the
GUI from the console.
However, the user won't see any messages neither,
when starting the exe from a graphical icon
on his desktop.

> (3) I like the Help->About page. However, if you go to Help->Contents,
>most of the links in the html page do not work. It seems to me that all
the >`external' pages do not work. For example, if I want to go to the
EChem++ >web page (line 2 of about), the link is wrong:
> >/home/bs/software/EChem++/documentation/html/http:/sourceforge.net/pro=
jects/echempp
>is certainly not correct. Of course, `http://...' etc. is the link in
the >EChem++ documentation page, but here this should not be appended
to the >EPP_HELPSRC path!
>
> (4) In the Help->Contents menu, the navigation tools work nicely.
However, >File->Exit does not work (nothing happens).

work in progress... there are a lot of other details
that do not work by now within the help browser.
However, the file->exit should work with the actual
cvs version.

> (5) If I enlarge the main window to full screen size, the numbers
>recording the window size in the left hand menu start to oscillate, the
>program uses up not really all, but a lot of the resources of the
computer, >and even after decreasing the size the same numbers (in my
case 1217x881 >pixels) remain in the left hand menu. Oszillation is
between 1217 and 1218. >this is true, regardless of a graph being
present on the screen or not.

I cannot reproduce this. Please check if this still
occurs after a fresh installation and with the
<prefix>/bin/model program.

> (6) In the main menu, at the beginning, under the heading Edit, both
>entries, `remove ...' and `deactivate ...' are active, although no
graph is >present yet. This should, I think, only be activated after a
graph is >displayed in the main window.

ok.

>
> (7) In a similar way, File->Save should only be active if some data are
>already present.

ok.

> (8) Same for File->print and File->export

ok.

> (9) Under heading Definitions at the beginning only Geometry is active.
>However, isn't mechanism independent of the geometry, and it should
make no >difference if the user starts with mechanism definition and
then defines >`geometry' or vice versa? Maybe there is a technical
reason, but for the >user this is not obvious.

the mechanism depends on the boundary indices wthin the
model geometry, since all heterogeneous chemcical
reactions must be linked to these boundaries.
(will be included in the user-doc)

> (10) By the way: what we define under heading Definitions->Geometry is
not >only the geometry but also some additional things, e.g. the
> excitation function, which has not really something to do with the
>geometry (the only link is that the excitation function has to be bound
to >a boundary ...). In the light of the discussion between me and Kai,
the >main heading here could be Definitions->Experiment, and then we
could >accomodate the various definitions of the excitation function,
the
>   geometry, and maybe in some future version even the collection of som=
e
> more or less complex experiments.

right - the dialog is somehow deprecated. I'll change
the name to "Experiment".

> (11) The selection menu CV/CS/UD should maybe read `user defined' for
the >third entry. There is enough space to write this, and UD is
> certainly not an obvious acronym for the novice user.

ok.

> (13) If I open the Definitions->Mechanism dialog, I see a default
>mechnaism which is rather complex. Could we have here a VERY simple
case, >if we want to have something at all? We could also provide users
with a >couple of example files, which they can load in the beginning to
have >example formulations in contrast to heving this rather arbitrary
mechanism.

ok. the example is due to some old testing of the compiler.

> (14) If we start the Definitions->Model parameters dialog, some entries
>contain default values, some don't. This is misleading for the novice.
>Either all or none of the entries should have defaults, or there should
be >at least a consistent pattern.
>
> (15) Unfortunately, my first experience of defining an exc.fct. etc.
>resulted in an excitation function error: Out of range, when trying to
>simulate. I used the mirror and the repeat function ... ;-(

did you press "append" after mirror? check for the
scan number in the upper right corner. Alain is currently
preparing an user docu - that may help.

> (16) OK, the second approach (only one scan in CV, no repeat or mirror)
>works fine. :-) There are a couple of inconsistencies in the
> concentration profile graph, but I can't repeat them, because after
>increasing this window to full size, I again have a frozen application.

not reproducible.

> (17) Simulating a CV (simple one electron ET, almost reversible) with
both >concentrations of A and B equal to 0.001 M  (0 -> 0.8 V; E0 =3D 0.5
V) causes >heavy oscillations at the beginning. This is probably a
numerical problem, >which could be avoided by using other solver
settings (would I say), but >for the novice again, this looks a little
> disappointing. Let's provide people with a couple of `good' examples
which >work.

that is indeed a numerical/modeling problem comparable to CA
simulations.
The product is initially present and is reduced immediately
at the beginning of the experiment (=3D> inconsistent initial
values).
Do you know, how DigiSim behaves in such a situation?
As with CA experiments, you should choose a very small initial
time step in order to get rid of the oscillations. However,
as you know, by now, the CA simulations have not been tested
quantitatively!

Regards,
Kai

--=20
http://echempp.sourceforge.net

Kai Ludwig
Institut f=FCr Organische Chemie
Auf der Morgenstelle 18
72076 T=FCbingen
Mail: kai...@un...