| Author |
 Topic  |
|
|
JoyceKai
3 Posts |
 Posted - 19 Aug 2004 : 13:58:10
|
I am getting a few reports from people that use my software that it crashes really bad when the graph is complete. Because it happens so infrequently I am unable to duplicate the problem. When it crashes on XP it wants to send an error report to Microsoft, which to me means a GPF.
I am using the graphing package with VB6. When the graph is complete, I show a messagebox that says "Graph complete". When the error occurs there is a LONG pause after the graph is complete, during which the user can see the graph on the screen. Then it crashes before the messagebox is shown.
Is there any general advice on the types of things to look at in figuring out what is causing this?
I have had 2 previous problems that caused the graph to crash similar to this. In both cases, the action that caused the graph to crash could have occurred minutes, or even hours prior to the crash. I will describe these 2 situations just so you can get an idea of what I am asking for.
1. I use a 3rd party printing package called ActiveReports by Data Dynamics. If I do not properly unload all components of the Report after the user does a printout, the next time he does a graph, BOOM!! Are there other situations where I need to be extra careful about making sure stuff is unloaded before I do a graph?
2. (I believe this was described in the Quinn-Curtis documentation.)
WORKAROUND FOR WRSetXYGraph PROBLEM. WRSetXYGraph must be called before WRSetScrollGraph so it gets LDataRes = 1. Call it during program startup in case user accesses WRSetScrollGraph first
Joyce Kai |
|
|
quinncurtis
1164 Posts |
Posted - 19 Aug 2004 : 19:16:51
|
Sorry, but there is no information to work with in what you describe. If a crash is happending a long time after a graph is complete, then I would suspect the problem is the code that is exectuing after the graph finishes. What is that code doing ?
Things that could cause that type of crash would be trying to access a pointer or a handle that is no longer valid.
1. We do not have any confirmed reports from regarding 3rd party components that cause problems with the Real-Time Graphics Tools.
2. We do not have any record of recommending the WORKAROUND that you describe. If you have a reference or an e-mail describing it please forward it to us. |
 |
|
|
JoyceKai
3 Posts |
Posted - 20 Aug 2004 : 10:09:47
|
The software crashes immediately after the graph is complete. It seems like execution never returns from the graphing package. (It crashes before the WGClosePage statement is reached.) However, the action that the user took that caused this crash may have been done minutes or hours earlier, so it is very difficult to figure out what I need to guard against.
In example 1, the user does a printout, and the next time he does a graph, it crashes. I fixed this problem by making sure the printing package components are properly unloaded.
In example 2, I have an XY graph and a scrolling graph in my software. If I don't include the workaround when the program starts, then if the user chooses to do the scrolling graph first, when he does an XY graph, it crashes. (This problem was found when we had the 16 bit version of your software, but I believe it still occurs in the 32 bit version. It was fixed before I hired on, so I don't know how they figured out how to fix it.)
Now my XY graph is crashing for some 3rd reason. It is happening very infrequently (12 times in the last 2 weeks, and they probably did around 800 graphs) but it is very inconvenient for them and makes us look bad. They swear they have not done any printing, so it must be a new cause.
As for losing pointers... I do a call to WGIsGraphDescValid before doing any processing in the UpdateTrace routine. It couldn't just lose a handle in the middle of the routine, could it? My calls to WGSavePageMeta and WGClosePage seem to be AFTER the crash occurs.
Joyce Kai |
|
|
|
quinncurtis
1164 Posts |
Posted - 20 Aug 2004 : 10:54:02
|
You have been using the software for many years. Yet this specific problem has just now started ?
If this is a recent, new problem, what is unique to these installations?
Is this problem occuring on a single system, on multiple systems at the same customer site, or on multiple systems at multiple customer sites.
Can you reproduce the problem in your office?
Is the data generated automatically ? If invalid data is allowed to enter the datasets this could cause indeterminent problems down the line. In this case every data value should be checked for validity before it is allowed in a dataset or to determine a setting for a graph.
You should remove your workaround if it from the 16-bit version of the software, since it is not applicable to the current revision of the software.
|
|
|
|
JoyceKai
3 Posts |
Posted - 23 Aug 2004 : 11:07:02
|
The problem has been seen to occur on at least 4 computers at different sites. Probably it is occuring at more places, but they just restart and go on, and since the error does not happen again right away, they don't complain. The customer that is complaining is facing a deadline.
If we could figure out how to reproduce it, then I could almost certainly fix it. I have asked the customer (where the problem has occurred 12 times in 2 weeks) to look for patterns, but they cannot discern any.
It is difficult to know what is different about the installation because the program contains many lines of code. Also, since the problem occurs so infrequently, I am not entirely certain at what point it started, so, different from WHAT?
The data is pressure readings, for both X and Y components of the graph. We have previously had clipping problems (which you were VERY helpful with) so each reading is now checked to make sure it is within the graph window.
I believe the workaround I mentioned IS applicable to the 32 bit version. I think I tried removing it when I switched to 32 bit and the graph crashed, so I put it back in.
Joyce Kai |
|
|
|
quinncurtis
1164 Posts |
Posted - 23 Aug 2004 : 11:36:43
|
If you wish do give us exact details regarding the "fix" you reference we can look into whether or not it is still needed. Our e-mail records go back 8 years and we have no reference to it. Anything associated with the 16- bit version could be 10 years or older.
Otherwise we have no course of action that we can recommend until we have a better idea of how to reproduce the problem. |
|
|
| |
Topic |
|
|
|
graph crashes
