Colour Touchscreen Crashing

Progress report on the CTC crash problem:

Firstly, the development was not outsourced. The software does make use of a well known operating system, a large amount of CIS developed software, and some 3rd party software components.

As to progress:

There has been a huge amount of effort gone into trying to figure out what the cause is. We have 10 touchscreens running a soak test to try and reproduce this. We have also been running development versions for a very long time without seeing the defect appear.

During the investigation, we have found that the crash occurs more often with some projects when compared to others. The project size seems to be a factor.

Initially the symptoms led us to believe the fault lay in our software. After a couple of weeks of digging through our code, and then through some very obscure parts of the Borland Delphi run-time system, we found that not to be the case.

Further investigation has found a memory leak in a portion of 3rd party software. We have gathered comprehensive evidence, and sent this to the manufacturer. We have been chasing them regularly to find what they are doing. So far the reaction time has been less than impressive.

At the moment we have a "patch" coming for testing tomorrow (thats 3rd Nov), but we don't know if this will be a fix or not.

---------

As a work-around, we have developed an update for PICED that will allow the CTC to be restarted by a control button on the screen, or from a schedule. This will allow the CTC to be regularly restarted, automatically if desired.

This update is currently being tested by the CIS test group. We hope to have it available very soon.

This is a blatant work-around which which is not a long-term solution, but it will relieve some of the pressure that everybody is under.

When this is available we will post to let you know.

 

A further progress report on the "Crashing Colour Touchscreen":

This has had a gut-busting effort for the lasty month (I'm not counting the time spent before that - which was also substantial).

There is light at the end of the tunnel.

I cannot yet give a definite date when the fix will be available (see below), but PRELIMINARY engineering development is looking good.

The next steps are:

1. Put back the full feature set of the current Operating System build (some has been hacked out for expediency).

2. Put in the latest service packs and patches from the Operating System vendor.

3. Make sure everything works. Update our various control / setup programs to suit if needed.

4. Open up a range of IP ports for the socket / logic developers to use.

5. Put it through the test group so that it can be independently hammered - so when it comes out it will not just be a good release, it will be damn good.

An initial GUESS is that there is probably 2-4 weeks work in the above list. The Christmas silly season is pretty much upon us, this will add another 1-2 weeks. So a ROUGH GUESS is that an update should be out around the end of January.

CIS apologies for the delay - but when it comes out we do want it to be really good.

 

Not immediately :-(

The priority is to have something that is nice and stable.

HTML will be a future update to the version of HomeGate that runs in the CTC.

 

Hi Potemkin

Your solution will help - but it only defers the inevitable crash.

 

A further update...

The new version should hopefully go to the test group in about a week.

The updated O/S is looking good, the firewall is working fine, we have modified the network setup utility so you can selectively open some ports (for those who want to use TCP/IP sockets)....

There is still a small HomeGate issue to sort out and I hope that will be fixed RSN (real soon now). Once that is done it can go to the formal test group who will bash the hell out of it. Then to release.

IN THE MEANTIME...

Use the scheduled restart facility. Its a very brute-force method, but the screens I have used it on only make a very quite "click" when they restart - set it for 2am or something like that - maybe once a week. This is a blatant workaround...