Question About Segment Errors

[Steve Albin]

I'll add my bit on this and you can take it for what it is 
worth.  I develop as a "hobbyist" and do not have a large user 
base, so I don't have vast experience.  I used to encounter 
these segment errors occasionally with VText fields.  I used 
them to allow users to add free form notes to a record.  I was 
not checking the return code on the updates to the DB 
(amateurish, I know) so I would not see when the errors were 
occurring.  I would only see the errors on doing a diagnose.  I 
suspect my users were making the update and then the update 
would "disappear" and they would then re-enter new notes.  I can 
only surmise.  Again, this is not a commercial app so it is only 
my reputation among friends and not my livelihood that I have to 
worry about.

The point is, these problems went away when I changed the 
application to update the VText fields separately.  I used to 
create a cursor something like "SELECT * FROM myTable" which 
would obviously include the VText fields and then update 
whichever fields needed updating.  I since changed to only 
select the non-VText fields and update those.  If a notes field 
is changing, then I create a separate cursor containing only 
that field and update or add separately.

I'm not saying this is the solution, but it seemed to solve my 
problem.  It may just be I'm just lucky though.  I am using V4RB 
4.3 now and I had seen these problems as far back as V4RB 2.

I hope this adds some light to the discussion.

Steve

--
Steve Albin - Montclair, NJ  - USA
http://www.jazzdiscography.com/
http://www.steve-albin.com/


On 5/7/10 at 11:26 AM, dev at infinitydatasystems.com (Infinity 
Development Team) wrote:

>======================Original Email====================
>
>
>>
>>>Believe me, if I could I'd report it to Mantis with examples. This is
>>> obviously a very subtle bug. But with lots of users it happens often
>>> enough that database segment errors are the #1 tech support issue I
>>> have to deal with.
>>
>>Just in scenario you described looks very simple to
>>
>>A) remember what exactly was made step by step
>>B) try reproduce it
>>
>>Its hard to be simpler..
>>
>>So this user self cannot reproduce now ?
>
>Yes, of course. I meant to imply that I *did* try to reproduce 
>it. And of course I couldn't. Nor can the user. But it 
>happened. As I said, it's a subtle bug. If I could reproduce 
>how to create segment errors of course I would report it.
>
>Jon
>=====================End Original Email===================
>
>
>
>My situation is very similar. I was using Valentina for an 
>email client. Consistently, when downloading email to a certain 
>mailbox, it would corrupt and throw up segment errors. I had a 
>backup copy of the mailbox that was perfectly fine, and 
>produced no errors. When the mailbox would get corrupted, I 
>would restore from the backup, and then try redownloading 
>(since I have my email set to not delete for a lengthy period 
>of time). Inevitiably, the fresh mailbox would always corrupt 
>the next time I would download email to it.
>
>The emails in question are a developer mailing mailing list 
>where all the emails are text only, not HTML part. The fields 
>that were producing segment errors were vText fields. These 
>fields, because they held the body and headers, actually had a 
>rather large segment size, at 2048.
>
>Unfortunately, as much as I love how fast Valentina is, I've 
>had to switch back to SQLite. I'm already behind on a deadline, 
>and spent too much time trying to figure this issue out. Once 
>my deadline is over though, I'll gladly spend some time 
>troubleshooting the above scenario and producing a working example.
>
>Matt
>_______________________________________________
>Valentina mailing list
>Valentina at lists.macserve.net
>http://lists.macserve.net/mailman/listinfo/valentina