Bug 84658

Created attachment 107316 [details]
Database crash messages

Would it be possible you attach the database file? (have in mind that any attachment are made automatically public so remove any private/confidential part)

Created attachment 107332 [details]
My odb file

Created attachment 107334 [details]
console bt

On pc Debian x86-64 with master sources updated yesterday, I could reproduce this.

I attached a bt

Lionel: bt shows a problem with iterators in baccess/source/core/api/RowSetCache.cxx, thought you might be interested in this one.

Cannot reproduce on master of 26 sep, Debian GNU/Linux amd64. Could someone please give more detailed reproduction instructions? Thanks.

I open form "Καταχώρηση ΝΟΜΟΥ".
Click on "new record".
Enter "74" into "A/A".
Click in "next record".

Works for me.

(In reply to Lionel Elie Mamane from comment #6)
> Cannot reproduce on master of 26 sep, Debian GNU/Linux amd64. Could someone
> please give more detailed reproduction instructions? Thanks.
> 
> I open form "Καταχώρηση ΝΟΜΟΥ".
> Click on "new record".
> Enter "74" into "A/A".
> Click in "next record".
> 
> Works for me.

Hello,

open form "Καταχώρηση ΠΟΛ"
click the button "new record" in the same height as blue and red button
base crashes

Indeed, I don't reproduce the problem with the first form but with the second one.

OK, reproduced with form "Καταχώρηση ΠΟΛ".

The crash is linked to subform "Νέο Άρθρο Νόμου"; if one removes all forms except "Κύρια" and "Νέο Άρθρο Νόμου", then it still crashes. If one removes "Νέο Άρθρο Νόμου" then it does not crash anymore.

"Νέο Άρθρο Νόμου" is a subform of form "Κύρια" *but* has no master/slave fields, meaning it is actually independent and not a subform. I moved "Νέο Άρθρο Νόμου" to be a "sister" of "Κύρια" instead of a daughter (a top-level form instead of a subform) and it does not crash.

(For full clarity, I agree it should not crash. Just documenting I found what is the exact condition that makes it crash.)

The crash is also not specific to "new record". It happens with *any* movement (next record, jump to record 50, jump to record 20, ...)

(In reply to Lionel Elie Mamane from comment #9)
> OK, reproduced with form "Καταχώρηση ΠΟΛ".
> 
> The crash is linked to subform "Νέο Άρθρο Νόμου"; if one removes all forms
> except "Κύρια" and "Νέο Άρθρο Νόμου", then it still crashes. If one removes
> "Νέο Άρθρο Νόμου" then it does not crash anymore.
> 
> "Νέο Άρθρο Νόμου" is a subform of form "Κύρια" *but* has no master/slave
> fields, meaning it is actually independent and not a subform. I moved "Νέο
> Άρθρο Νόμου" to be a "sister" of "Κύρια" instead of a daughter (a top-level
> form instead of a subform) and it does not crash.
> 
> (For full clarity, I agree it should not crash. Just documenting I found
> what is the exact condition that makes it crash.)

Hello,

thanks for your help.

I changed the subforms "Νέος νόμος", "Νέο Άρθρο Νόμου" and "Νέα ετικέτα" to main forms, because they are not subforms to the "Κύρια" form. Their use is to insert only new records to the connected tables.

Now i can insert new records to the main (Κύρια) form. It works.

Crash is also linked to the "new data only" property being set. It is because the RowSet is reexecuted while in insert mode, so the rowsetcache is reset with a new rowset, *but* the rowset is not anymore in insert mode, but the iterator to the rowsetcache (m_aCurrentRow) still points to the rowsetcache's insert row.

Created attachment 107415 [details]
minimised reproduction case

http://cgit.freedesktop.org/libreoffice/core/commit/?id=195674627cb887b1e0d92d756a285ad48ef1d362