Mail Me

Week 40, 1999 ,Svenson

Sjon




-- Links -- Calendar -- Jobs to do --
-- Previous week -- Today -- Next week --


This is the past


1999-10-04

top
Mon
Tue
Wed
Thu
Fri
Sat
Sun
bottom

Monday

We are back to double figure temperatures (10C) accompanied by some short but serious bursts of rain.


The Americans can be exasperating at times (see the past few weeks) but they can be nice as well.

They installed TeleSales on the production environment, swapping the sales call centre over from their old system. The only problem that cropped up was due to some of the local NT Workstation machines being configured for 'English (international)' and others for 'English (American)'. The result is that some have a decimal point and others have a decimal comma. This is only confusing for the operator but causes no errors in processing. To get things right they must uninstall TeleSales from those PC's, change the NT setting and reinstall TeleSales. They are going to do that either this evening or next weekend. That was the only problem noted.

Because things went so well the sent a mail thanking us. Now that is something we are not used to over here.
Either it works as it should so why thank anyone for doing what he should do. Or else it doesn't work and then you have to complain to get things repaired.


Time (way over time actually) to do some bills and get the balance sorted out. I do this by hand, then I know exactly who to blame :-/


 


1999-10-05

top
Mon
Tue
Wed
Thu
Fri
Sat
Sun
bottom

Tuesday

We had a load of cold (9C) rain in the morning and a few spells of drizzle during the day. Most of the time it stayed dray and even sunny.


Almost exactly one year ago (1998/10/01) the documentation for TeleSales was finished. It was then pulled trough a resource compiler to create Windows help files that were (manually) linked to the application. I did not yet work on TeleSales then. Since then nothing has been done to the documentation so the user manual and the online help are hopelessly out of sync with the application.

We got news today that a documentalist will be assigned to the project for three weeks to get the manual and the help redone. The person doesn't know anything about our application so we have to train him first and the intention is that we must keep the manual and help up to date afterwards so he must teach us how to accomplish that. Three weeks is of course ridiculously short. We spent the better part of the day just tracking changes in the screens compared to the screen prints in the current manual resulting in a 7 page listing of screen names. The best we can hope for is that the most used parts will be covered. Not because they are the most needed, quite the reverse is true a manual is important for things you don't do often enough to remember, but because that will be the part that is most likely to be checked.
Hey this looks like the documentation process used by Microsoft.

Before any one of the manual writers among the day notes gang smell an opportunity here, let me warn you, more than about 15 or 20 books, however good, will never be sold.

 

And the French reported a problem (via OLI) about the possibility to add phased out items on orders causing problems in the logistics module. They claim to do that by putting the phased out item on the second (or third or ..) line of the order. The first line traps the error. I tried to reproduce that situation locally but without success. I simply could not get a phased out item ordered unless there was still stock available, in which case there was no problem with logistics. I will have to check the on their system tomorrow because I don't have a user profile to log on with now. Wonder what they are doing different.

 

And there is, at last, a problem in America. It is not a problem of our making though. The logistics team has changed some common routines and added a server module. This was sent to America, amongst others, on Thursday. Some of the changed routines are used to convert commercial orders (TeleSales, OMSI3) into logistic orders (OLI). The Americans don't use OLI but a totally different product for their logistics so they changed the conversion routines, based on the old version, to work as a pass trough between TeleSales and their own logistics module. Because the new version had to do with OLI, a module they don't use, and because it came so late the Americans did not apply the changes to their conversion programs. On Friday we shipped a quick update to solve a totally unrelated problem but containing one of the affected programs.

With a little bit of assistance they found out where the problem was and they will correct things by applying the changes. That is what we call 'panic football' here in Belgium (for the Americans that translates into 'panic soccer' of course).


That is what one call an average day on the job. I didn't even have time to read more than Toms and Shaws day notes.

 

I got a reminder to pay an invoice that I already paid. They probably use a computer do the invoicing. So I am busy with paperwork.


 

 


1999-10-06

top
Mon
Tue
Wed
Thu
Fri
Sat
Sun
bottom

Wednesday

We had two or three light showers during the day but calm sunny weather mainly. The temperature however had to climb from a low of 5C this morning to just 17C in the afternoon. That is fine for me but most people start to lament about these 'very cold temperatures'. I just smile at them and mutter something like ' well just wait till it actually gets cold'.

Now, I wonder, why don't they smile back :-P


The Americans experienced a bunch of 'socket errors' on their clients yesterday. They want us to look into that and they want us to push Progress for a solution. At first they blamed us for it, and Peter is inclined to give in to them.
When a socket error occurs the user of that specific client has to restart TeleSales and he has to re-enter the order he was creating.
In the joblogs on the AS/400 we find a lot of cancelled jobs with TCP error xxxx. This error does not occur at a specific point or after a fixed series of actions. It looks random. The only way we can reliably reproduce the message in the joblog of a server is by killing a client process either via task manager or via the reset button or by pulling the network cable. To get a 'socket error' on a client we have to end the server job abnormally.
All this points rather sharply to a network problem but we have no access to their NT servers and even if we had we don't know enough about network logs to make conclusions. And we even have a hard time convincing them to have a look for themselves. They have people there that know enough about networks. The only one here that knows something about NT and networks is Wilbert but he is not in today and he has the habit of diagnosing this type of problem on his own, he is not likely to ask the Americans for help. This takes time which we don't really have. He has communicated the problem back with Progress but he is not going to push to hard for a solution.

We got a mail from the business manager from America asking for a solution fast. I sent a mail back, to our contact person there, telling him what we want them to look for and why we think they should do it. I hope this gets something moving because if things keep crashing they may decide to revert back to the old system and that could break the whole TeleSales project.

So I spent almost the whole day reading joblogs and counting errors and time intervals and other really joyful stuff :-(


I am experimenting with a new day note system (check the today link on top of the page) and I noticed that a black background is far less objectable if you are using a good LCD screen. It is not better than a light background but on a LCD screen the glowing of the white text is markedly less intense. Now if there is a page with black background where the text is too small or the glowing to strong the easiest way to make things readable again is by simply selecting the text. This reverses the text back to black on white. It is not perfect because you loose colour which makes links less visible and when you accidentally click somewhere the text reverses again but for a small paragraph or a caption it is a solution.

An old DSTN type LCD will not do because ghosting aclually blurrs the text. You need a good TFT LCD, something like the LG 500LC that I am using now for example :-§


 


1999-10-07

top
Mon
Tue
Wed
Thu
Fri
Sat
Sun
bottom

Thursday

It was a generally nice day; weather wise that is. Starting off at 6h00 with a cool 7C but rising fast and going back down to 16C at around 17h30. No rain at all and lots of sun.


We had no new problems from America.
The 'socket error' problem did not occur as often as the previous day. I counted 18 instances but quite some were duplicates, that is the user gets one crash but the application reacts by automatically restarting, and crashing again. On the AS/400 that results in two crashed jobs but if you check the timing and IP addresses in the logs you can combine some logs to form such a double crash. Peter called America and they have started some kind of 'sniffer' on the network to look for potential lethal events. I guess we will get a report tomorrow or at least by next week.
And the people from Progress seem to have been kicked into action. Wilbert is back and he sent a mail to Progress containing the mail that I sent to America. Maybe that's what convinced them we need a solution.

All in all I again lost some hours slogging trough joblogs and assisting Wilbert. Lovely :(

 

I did have time for the French problem. And it is one of those mystery problems all programmers love. Sven, from OLI had given me a profile for logging in on the French machine and ha also gave me a few item numbers and addresses. I had copied the item number wrongly but I found a good phased out item after some querying. I made a few orders and the phased out item was trapped each time. So I couldn't reproduce the problem there and I could not reproduce the problem here. Who is fooling who here? I got the manager in France on the phone and he guided me trough the process, with another phased out item and lo, it passed all the test and resulted in a logistics order. After some searching we found that there was still some stock for that item so it should pass as it did. No problem.

Just to make sure, I try to be thorough, we removed all the stock for the item. This time it should generate an error, preventing the order from being released. Only it didn't. It just passed on to logistics as if, well just like that. So we do have a problem.
Some items behave as they should, some sneak trough depending on the position they are in.

Debugging remotely was out of the question so I checked and rechecked the code and I munched the data. All to no avail. As a last resort I ran the data trough a virtual computer (that is that wet grey stuff called a brain) stepping trough the lines of code on by one. I did find a problem that may cause the symptoms we are seeing. Actually it is the coincidence of two independent errors. I will correct that tomorrow an let the French test it.

 

And then Peter came around with two new adaptations to include in OMSI-3. I am not yet out of work.


I have been having problems with connecting to the ISP here. The problem is that they had a very good and well performing service. So many new clients got attracted causing the performance to drop. In the first four months I never got a busy signal and the line was never dropped. The last month (September) I often got dropped lines and gradually I got more and more busy signals. Every day of the last week I had to try several time to get a connection. Yesterday I tried about 10 times between 21h00 and 24h00 and I never got a connection.
So if you missed my day note, I am not getting lazy, I am just not getting a connection..

I tried today about 5 times between 19h40 and 21h00 without success. Maybe I cannot post today either. I am not feeling happy about this but there is not much I can do at the moment.


 


1999-10-08

top
Mon
Tue
Wed
Thu
Fri
Sat
Sun
bottom

Friday

10C and dry in the morning but just a tad windy. It started raining at noon till, well, tomorrow some when.


The French problem is tricky because some phased out items produce the problem and others are handled correctly. I found a double bug in our order entry program yesterday but although it should produce the results we see it should not make a distinction between different items and I should be capable of reproducing it locally. So even while it is obviously a bug it is unlikely to be the cause of the reported problem.
So I restored the program with bug (I saved the corrected source) and started wrestling with it. At the end of the day I am 1-0 down. There is more to this program than meets the eye.

We did get a few snippets of information from America but nothing conclusive. There is a trend but not yet enough data to sure about it. The socket error occurs most on a series of older computers. Without the actual data we don't know what that actually means. What is most? What do they call older computers? How many computers are that?
Too many questions, too little figures. But we keep looking and counting.


Tomorrow we are going to replace the gate of the garage. Because the weather will be wet next week (well that is the forecast) this will have to be done in one go. The old gate is in such a condition that once removed it cannot be placed back so for this job failure is no option. The new gate must be placed and must fit.

So expect tomorrow's day note to be short.

The update Yesterday was problematic again. I tried to connect about 20 times between 19h40 and 23h40. I always got a busy line. That is somewhat later than my normal bed time, I have to get up at 5h45 and I have to actually sleep about five hours. I cannot afford to nap on the job and sleeping in my car could be deathly. So at 23h40 I tried one last time and with success. I had in fact a very snappy connection.

So what is it going to be today?


 


1999-10-09

top
Mon
Tue
Wed
Thu
Fri
Sat
Sun
bottom

Saturday

Overcast, with occasional raining and consequentially warm in the morning (11C) and cool by day (17C).


I have been working on the gate with my brother, from about 8h00 until 23h40. With just two pauses for refuelling, I made a huge bows of rice with a vegetable mix (peas, beans, carrots, etc. ), sliced mushrooms and bacon.


 


1999-10-10

top
Mon
Tue
Wed
Thu
Fri
Sat
Sun
bottom

Sunday

I don't know if I should call this light rain or heavy mist, everything gets wet regardless. Thanks to the cloud cover its is just a bit too warm for running (14C).


We had the gate installed yesterday but there was more than enough job left to keep us busy till 21h00. Apart from three doors connected by hinges there are bolts and locks and draught list and ... Well about a million small things that must be done.


While moving tools beck to my brother I had a look at Cindy. Peter said there was still no sound. When I fixed her I tested and got sound. Weird. I start her up and indeed there is no sound I open her up check the cables but everything looks normal. When I tried to change the properties of the Soundblaster card I got a message about the device being in use by some other process. What process? I just started the PC with nothing in the Start-up folder, so there is nothing actually running. I re-plug the boxes in the green socket (I tried the others as a matter of routine) and hey presto! We got sound.

I don't know what was wrong but now it works. So I quickly close the box and run :-)


 

-- Top -- Bottom --

-- Links -- calendar -- Previous week -- Next week --


Svenson © 1999

A day you don't learn something new is a wasted day.