| Author |
 Topic  |
|
|
Luke1410
Senior Member
32 Posts |
 Posted - May 06 2008 : 4:23:22 PM
|
Hi,
I'm testing the latest beta build with our code base (C++) and had some trouble. After using it in VS 2008 for several hours I ended up with the "Go to definition / declaration" features stop to working (couldn't find the corresponding definition/declaration) as did VA's own "Find References" feature which displayed the error message: "Please place the caret on a symbol before running FindReferences.".
After rebuilding VA's symbol library everything worked once again.
I was wondering whether VA has some trouble parsing some of our code correctly. Therefore I enabled VA's logging feature and checked the log file for any suspicious entries and found a lot of entries like the following one:
"IsTemplateCls (1): Brace Mismatch )
{
if (!bCas"
Is VA having problems with some of our code constructs, which might lead to the problems described above, or can those log messages be safely ignored? |
Edited by - Luke1410 on May 06 2008 4:29:28 PM |
|
|
feline
Whole Tomato Software
United Kingdom
19021 Posts |
Posted - May 06 2008 : 5:01:08 PM
|
How are you triggering "Go to definition / declaration" ?
There are two items called this in the IDE right click menu, but these are IDE commands, and should not be effected one way or the other by VA.
The problem with Find References clearly shows a VA problem.
Are you seeing any odd underlining from VA? I am assuming that you have:
VA Options -> Advanced -> Underlines -> Underline mistyped symbols using
turned on. If our parser is getting confused, which sounds likely, then I would expect valid code to start being underlined as mistyped symbols. |
zen is the art of being at one with the two'ness |
 |
|
|
Luke1410
Senior Member
32 Posts |
Posted - May 06 2008 : 6:32:28 PM
|
I trigger the "Go to definition/declaration" stuff by rightclicking on an identifier (method/member variable).
@odd underlining: Yes, I did see a lot of underlined (but valid) identifiers. Furthermore the coloring of identifiers was "incorrect", too (i.e. identifiers which should have been green (local variables) or dark red (methods) were black. Last but not least when writing a "." after an identifier which was a pointer to a class, VA did neither automatically replace the "." with a "->" nor did it show a list of valid member methods. |
Edited by - Luke1410 on May 06 2008 6:32:41 PM |
|
|
|
feline
Whole Tomato Software
United Kingdom
19021 Posts |
Posted - May 07 2008 : 08:49:48 AM
|
I have no idea why the IDE's commands are having problems, but clearly something is confusing VA's parser in your code.
In a cpp file where you are seeing this problem, if you use Alt-m to get a list of functions in the file is this list correct? If the first few functions are listed, but a problem then shows up, this should give a clue as to what is causing the problem.
If you show VA Outline, does this show the correct information for your files? Or is this having problems as well?
Do you see problems in all files, or only some files?
I am looking for any clues to help explain the problems you are seeing. |
zen is the art of being at one with the two'ness |
|
|
|
Luke1410
Senior Member
32 Posts |
Posted - May 09 2008 : 07:14:47 AM
|
Unfortunately the problem I expected can't reliably be reproduced and since I rebuilt the symbol files, neither can I check what would have happened if I had used Alt-m nor how the VA Outline looked like.
If I remember correctly, the problem was present in all source files (not only a particular one).
As I was asking in my initial post: Does VA provide any means to display problems it has with parsing our code which could either be used to:
- report the problems here
- adjust the code sections VA has trouble with to work properly? |
|
|
|
feline
Whole Tomato Software
United Kingdom
19021 Posts |
Posted - May 09 2008 : 09:16:22 AM
|
Sort of. It really depends on the type of problem you are seeing.
If you have code in the 6th function of a file that confuses VA's parser then looking at VA Outline or the Alt-m list should help, since they will be correct for the first 5 functions, and then the problem will show up.
Complex macros are a common source of problems in code, so it is worth looking for macros when problems show up. However given the massive range of things people can and do with complex C / C++ macros I cannot offer any general comments here.
Sometimes log files will help.
It all depends on the nature of the problem.
Most of the time a parser problem is file specific. It may effect most files in a solution, if it is triggered by a common include header file, but there are normally a few files that are not effected.
If the problems appear global then this may be caused by a problem with VA's symbol database. If rebuilding the symbol database fixes the problem then this is a good indicator that is where the problem lay. Our symbol database is not supposed to become corrupt, so tracking down the cause of the problem can be tricky.
We have made some changes for the next build of VA to try and make the symbol database more robust when working with multiple IDE's at the same time. |
zen is the art of being at one with the two'ness |
|
|
|
Luke1410
Senior Member
32 Posts |
Posted - May 12 2008 : 12:37:29 PM
|
Thanks for the detailed description. If the problem reappears, I'll take a look at the outline list to see if it's caused by something VA had problems parsing correctly.
Furthermore if I may post a feature request: Enterprise Architect from Sparx provides a nice feature to easily identify code its own parser has problems to parse correctly: They simply print out the code lines / files their parser couldn't parse correctly. That helps a lot with identifying troublesome code and report it to their support center. It not only saved us hours of debugging sessions but no doubt helps Sparx a lot to improve the quality of their own parser, since bug reports can be made easier and contain all necessary information to create a suitable test case. |
|
|
|
feline
Whole Tomato Software
United Kingdom
19021 Posts |
Posted - May 13 2008 : 09:25:18 AM
|
I really like the idea of our parser logging any code it had a problem with 
At the same time, VA is designed to scan code as you are editing it, so often the code is invalid, you are still writing it, so it does not always make sense.
*considers* if this only happened when logging was turned on, it should limit any problems due to editing...
I have put in a feature request to see what our developers make of this:
case=16650 |
zen is the art of being at one with the two'ness |
|
|
|
rittjc
Ketchup Master
USA
84 Posts |
Posted - May 13 2008 : 09:46:00 AM
|
quote:
Originally posted by feline
I really like the idea of our parser logging any code it had a problem with
At the same time, VA is designed to scan code as you are editing it, so often the code is invalid, you are still writing it, so it does not always make sense.
*considers* if this only happened when logging was turned on, it should limit any problems due to editing...
I have put in a feature request to see what our developers make of this:
case=16650
Please make this an option. Logging would be good if we have a problem, but I personally wouldn't want it to slow down normal operation, no matter how much benefit it brings to being able to resolve problems.
JMHO |
|
|
|
feline
Whole Tomato Software
United Kingdom
19021 Posts |
Posted - May 13 2008 : 10:23:02 AM
|
| I suspect this would only be done if VA logging was turned on, and you have to turn this on manually. Plus it turns its self off when you restart the IDE. |
zen is the art of being at one with the two'ness |
|
|
| |
Topic |
|
Go To Definition stops working with 1635
