OI 9.0.0 - Oinsight.exe crash (OpenInsight 32-bit Specific)
At 01 SEP 2009 06:59:45AM Simon Wilmot wrote:
Hi all,
I am having some issues with an Oinsight crash, happening in various guises, but always with an offset of 0005a0e2.
Can anyone point me in the direction of either a fix from RevTech, what I need to check in our system is set up correctly, or a mixture of both ??
TIA,
Simon
At 01 SEP 2009 09:04AM Jared Bratu wrote:
How frequently can you reproduce this problem? A OEPROFILE.LOG file in the root of the OI directory may help pin down when the crash occurs.
Is this system upgraded from 8.x or is this a fresh copy of OI?
It may also be worth trying to upgrade to 9.1 as there are several updates included in it. InList() had an intermittent bug that was fixed which could cause sporadic crashes.
At 01 SEP 2009 09:14AM Simon Wilmot wrote:
It can be done pretty much at will at the moment. I presume that I need to create the aforementioned log file for it to record into.
The system is upgraded from 8.0.0 to 8.0.3 and then to 9.0.0
I will look at installing 9.1 to see if it has any affect.
At 01 SEP 2009 11:03AM Simon Wilmot wrote:
To update, installing 9.1 appears initially to have resolved the crashes I was experiencing along with another one that had been reported. However, there is an issue (Bill Paisley has raised in the Bug Tracker) in that 9.1 introduces a new System routine of 'Banner', which causes issues with our current version of that program. This follows on with a similar issue with the routine 'Find' in a previous version (and Number many moons ago). Given the nature of the names of these routines, has it not occurred to RevTech programmers that users of their systems will already be using routines with these names and if they created new routines with what would be more unique names (eg RTI_ …) that is would cause users much less aggravation.
As a follow up to this, currently I have removed the above entries from SysprocNames in order to re-compile our original versions. Number has never created an issue, but can someone enlighten me on the use of 'Find', presumably obvious, and 'Banner'.
Thanks,
Simon
At 02 SEP 2009 08:44AM Mike Ruane wrote:
Simon-
As I check prior releases of OpenInsight, it appears to me that the $BANNER routine has been included in OpenInsight since at least version 7.2. It was originally written in 1990, and has been carried forward as a part of the Arev32 implementation.
In fact, it has occurred to RevTech programmers, and staff, and myself, that we should prefix 'our' or 'System' routines with something unique, and that is why there are nearly 150 items in the SYSOBJ file that start with the prefix '$RTI_'
We are keenly aware that developers often think in like terms and would name programs with similar names.
In fact, since all of RTI's routines are rolled out in the SYSPROG application, any application specific items should still run as they did before, unless developers are putting these items into the SYSPROG application.
Is the core issue here the names of the programs, or the listing of the names in the SYSPROCNAMES record?
And, it is always an option for individual developers or organizations to use their own prefixes.
Thanks-
Mike Ruane
Revelation Software
At 02 SEP 2009 09:40AM Simon Wilmot wrote:
Hi Mike,
Sorry for the rant !! It was just frustration … I think the issue here is just the inclusion of these names in the SysProcNames record.
How does the SysProcNames record affect things - as as soon as I had installed the 9.1 upgrade and accessed a window I had a debugger crash in Banner (OI version). From this I presume that if a program exists in the SysProcNames record then the system ignores 'local' copies of the object and only runs the SysProg versions. Currently I have deleted the 3 routines that knowlingly affect us, but from what Bill told me the other day the SysProcNames record has grown quite a bit from 8 to 9.1 (ignoring the glitch at 9.0).
Agreed on the own prefixes situation, but I am working with code that was generated many years ago.
Regards,
Simon
At 02 SEP 2009 10:04AM Sean FitzSimons wrote:
Simon,
You are correct in the SYSPROCNAMES functionality. The 9.0 glitch forced us to rebuild the record in 9.1 and, as you have seen, it is a much a bigger record. The deletion of the lines that are affected is also the best workaround.
It seems that a White Paper is warranted regarding the SYSPROCNAMES record and the necessary programs that should be listed within the record.
Thank you,
Sean FitzSimons
Revelation Software
At 02 SEP 2009 12:03PM [email protected] wrote:
… or you could just read this :)
The post was made after hitting naming collisions during a 4.13-]9.0 upgrade…
Battlestar Sprezzatura - BSG 77
Colonial leaders in all things RevSoft
At 02 SEP 2009 01:37PM Sean FitzSimons wrote:
Carl,
Aha! I have started a draft of a paper that seems very close to yours ( I placed the Compilation portion above the Program Load portion ). Oh, well!
Folks, listen to Carl, if you are not already.
Cheers,
Sean
At 02 SEP 2009 04:53PM Mike Ruane wrote:
'Listen to Carl'…
The last thing he said to me was 'go have a beer for me', so…
Mike
At 04 SEP 2009 04:35AM [email protected] wrote:
If only more folks listened to me the world would be a better place … :)
Battlestar Sprezzatura - BSG 77
Colonial leaders in all things RevSoft