It's now May.28. Today I finished the code to translate from multi-UI
structured representation to swing components, specifically:
converting button to translate from struct, implementing flowpane and
stackofpanes. Then I put all the pieces together to actually show such
a converted object on screen. Next I put more pieces together to yield
an instance method within ClientWriter (the output stream within each
WrappedHtmlServletResponse object) which produces a swing object from
the current contents of the ClientWriter. Next I updated the toplevel
test rig so that after it runs the servlet and shows the result
converted to s-expression and HTML it also displays it in swing.  So
then I ran the first part of the first use case, where the initial
connection is made to the servlet and it returns the login form, and
it works! Before continuing, I finalized yesterday's log and wrote
this text so-far in today's, now 15:10, and next I'll update the
1-README.html file ... done, and upload both to the JavaClass WebSite
...  done at 15:55.

After break, resumed work at 19:23. Upgrading DispatcherServlet as
follows: If invalid login, have response include the login form again
(after the explanation of failed first attempt at login) ... but first
I need to delete some of the logged in users so they can log back in,
but my tool for auditing the ActiveUsers table is giving trouble, says
the DB update failed, aha I had '' around the numeric contents, will
fix that to use number instead of quoted string number, done, but it's
still complaining that the update failed, recompiling with trace on,
tried again, Cloudscape is giving me this error:

DELETE FROM ActiveUsers WHERE UserID=5;
ERROR 25502: An SQL data change is not permitted for a read-only connection,
  user or database.

That's not supposed to happen. It never happened before. This was
working fine just before I was having trouble with the J2EE server so
badly that neither Cloudscape nor J2EE could shut down normally so I
had to shut down the whole Linux system and restart it. So maybe
Cloudscape is in some wedged state? I'm desperate! I shut down
Cloudscape, then restarted it, then re-tried the same operation, and
now this happens:

DELETE FROM ActiveUsers WHERE UserID=5;
1 row inserted/updated/deleted

Phew! If it hadn't gotten working, I'd be unable to make any progress
on the whole Lab#3 project until I get help next Tuesday or maybe
never. Anyway, that was manually issuing the command from cloudscape
-isql. Now let's get back to my ActiveUsers audit program and see if
that is back to working too, at 20:08 ... yes, that works too, phew!
Also I needed to fix my Users audit program and the shell script for
calling it, because toSexpr was moved from Lab3Props to Sexpr1, and
that file was moved from the DL3 directory to another. Anyway, all
fixed at 20:50, ready to get back to work on DispatcherServlet ...
much cleanup done before bedtime.


May.29 08:41 looking at my original use cases document for first time
in weeks, to see what to do next: I see that I've implemented the
logic wrong. The program is supposed to check for login-session token
first, before checking for program-task mode. In that way, if user is
already logged in, even if no program-task mode given, the user gets a
main menu instead of a login form. So first order of business is to
reorganize the main logic of DispatcherServlet to look at both
jsessionid in the request properties (from hidden field or url) and
the ID of the session object, to see whether either is valid (matches
current record in ActiveUsers), installed new method in wrapped
request to pass through requested ID, put code in DispatcherServlet to
look at three sources of ID (session object, requested ID, and hidden
field in form), writing code to validate any that are non-null, but
first need to upgrade MyInput to support default values ... done and
revived&upgraded test suite for MyInput, done at 10:58. Now back to
test rig for upgraded audit of ActiveUsers to support lookup by
session ID ...

Had to stop for a while to run other urgent errands, back at 15:07,
back to working on new code and test in Lab3Props ... all done at
15:46. Now back to writing code in ActiveUsers to lookup session ID
and return record number, which uses that Lab3Props code, and which is
used by ActiveUsers audit where I'm testing it ...  finished and
working at 15:57. Now back to code in DispatcherServlet which needs to
look up a session ID in order to validate it to determine whether the
current user is already logged in ... taking nap at 17:20 ...

After nap and some other necessary tasks, back to this at 19:50
... almost done May.30 at 00:44, all parts of use case #1 (check
session and possibly log in) done except where session ID was lost and
user tries to log in when already logged in. In cases where user is
already logged in and passes correct session ID token, or where user
newly logs in, user gets toplevel menu. Cleaned up some stuff, now at
00:59 I'm starting up the J2EE server for the first time all day, to
see if a live run of everything so-far actually works ... while it's
starting up I remembered to copy all the recently-changed class files
to the system directory. Now starting up lynx ... done, at 01:09 still
waiting for J2EE server to be finish getting up ... server finally up
at 01:13, clicked on link ... at 01:17 I'm still waiting for JSP to
compile and server to offer me a cookie ... at 01:21 I decided to
check the system log to see what's going on: Program is running, doing
database query to see whether session ID from live J2EE servlet
container matches any session ID of logged in taxicab-message user ...
at 01:26 still no cookie, so going to check system log again ... it
seems to still be in that SQL query, no more trace output since it
showed me the SQL command ... finally at 01:31 it finished the SQL
query and moved on with the rest of the program and offered a cookie,
and I accepted, and it put up login form, and I filled it out, and at
01:33 I'm waiting for the login verification to finish. Going to check
system log now at 01:35 ... as I suspected, it's sitting in the SQL
query for session ID again ... at 01:41 it finished that SQL query and
moved ahead to start checking the contents of the login form that I
had submitted ... at 01:45 it's finally finished with those database
checks, and before I can switch back to the lynx window it's already
put up the main menu. I clicked on the button to send a message, which
feature is not yet implemented, so it should bomb out as soon as it
gets to that point in the program.

Bottom line: My method of developing program in NULL test rig,
augmented with conversion of output to swing, never having to do
anything with J2EE server until I have it all working, is a big win!

OK, at 01:50, server error: Unknown case of sTaskval [sendmsg], as
expected. Let me press back button on browser and make it die at
logout likewise ... did that, now 01:52 ... at 01:56 the expected
server error: Unknown case of sTaskval [logout].


Now 2005.May.30 at 08:16. Adding third button (show report of messages
per cab) to main menu ... done. Fixing login logic so that if user
tries to log in when already logged in, the old session ID will be
retrieved and used instead of rejecting the login ... done.  I ran
some other errands during that time. It's now 11:10 at the time I
claim that use case #1 is fully implmented, and tested in NULL test
rig, including swing display. (I don't feel like spending two hours
just to verify it also works in live J2EE session.)

Next to start working on use case #2, send message ... took nap etc.,
back to work at 13:01 ... made a nifty demo of the locking mechanism
whereby a program tries to get a lock, keeps trying until it gets it,
then holds onto it for 30 seconds before releasing, and when I start
two of them nearly simultaneously one of them gets the lock and holds
onto it while the other keeps trying, then as soon as the first one
releases it the other gets it promptly. Now working on code to
actually look up the largest existing message number and add one and
write a new message with that unique message number while holding onto
the lock ... done and tested at 15:38 but with one problem: When I ran
two simultaneous programs, after the first program got the lock and
posted a message and released the lock, the second was still unable to
get the lock. cloudscape -isql confirms the lock is no longer present,
so why can't the second program get the lock?? I changed the program
to show what's happening at the SQL level when trying to set a lock,
and in one case it got a truncation error because Locks allows only 10
characters in a user name and I happened to specify a longer name
during that test. Maybe that is what happened the other time it could
never get a lock, but that's rolled off-screen so there's no way to
see if that was the case. All this is because I originally created the
Locks table manually in cloudscape -isql. I've added code in
Locks.java to create the table with 20 characters allowed in a
username, same as in the Users table. So now I need to call that code
from somewhere. I think it's time to write a 'Setup' module which
makes sure all the tables are present and creates them if not. I'll do
that after nap&shower. It's now 16:23.

Resuming now at 17:30. Starting to write Setup.java ... got the first
part all done where it checks system configuration against properties
file to verify it's running on the correct system, and then verifies
it can open the database connection. Next it checks each of the tables
to make sure it exists and has initial data if any is available, which
requires calling mayMakeTable() within each specific class dealing
with the corresponding table. I needed to upgrade Users.java, add code
to Locks.java to create the table if it didn't already exist, which it
didn't because I had to delete it earlier tonight because the
manually-generated version I'd been using was bad as described above,
create Msgs.java from scratch, including code to create the table if
it didn't exist and put one dummy message in it, create Cabs.java from
scratch, including code to create the table if it didn't exist, and
set up initial data in properties file and software to load that
initial data. It's already 22:19, more than two hours past the
deadline I had set for myself to finish this lab assignment and upload
it in case I don't have time tomorrow, and I still need to create
MCRelat.java like the other table-specific classes already done, and
at the moment DispatcherServlet is in the middle of an edit so it
can't be compiled, and I still haven't implemented logout in the
servlet, and I haven't implemented either of the "interesting" use
cases (send message, show table of per-cab message counts), so I can't
stop work now and upload what I have so-far. I need more time after
Tuesday!!