[Kmymoney-devel] CSV Importer - A new problem, crash on import

[Chris]

About the email. I have checked my logs, apart from the error when I tried 
to send the "setenv.cmd" file I have no other non deliveries to you. Perhaps 
something at your end. I know hotmail is blocking me for some unknown 
reason. but then they have the crappiest spam control system of all the big 
providers, and I have given up fighting with it. Generally the only mail I 
ever get to my hotmail account is spam. so I assume they only deliver spam 
and block all genuine email.

As for the application

I changed the line
640: m_startLine = profilesGroup.readEntry("StartLine", -1) + 1;
to
640: m_startLine = profilesGroup.readEntry("StartLine", 0) + 1;

I realise a dropdown with a -1 value has no item selected. but even if that 
is the desired outcome before passing it through to the table indexing 
function it needs to be validated otherwise if the user fails to select a 
starting line the same error is going to occur. I assume this is checked but 
haven't verified it myself.


Chris


On 15/04/2014 9:35 PM, Allan wrote:
> On 15/04/14 12:12, Chris wrote:
>> Should have read...
>>
>> At line line 1517 it is used to initialise '*m_topLine*' which in turn
>> is used to initialise 'row' at line 1546
>>
>> Chris
>
> Hi Chris
>
> I'll look into this, but just to let you know I didn't get your first 
> email.  In fact, this one I did get is probably the one to the list.
>
> Just briefly, for now, the value -1 is used to obtain an empty combobox 
> field.
>
> I know in several places I do check for an empty cell producing a null 
> pointer, but will look specifically at this.
>
> Apart from that, it may be that it depends upon your file content, If you 
> could provide an  edited version which shows the problem. Strange that 
> it's OK on Kubuntu.  I assume both are based on the same source rev?
>
> Allan
>
>
>>
>> On 15/04/2014 8:23 PM, Chris wrote:
>>> Hi Allan
>>>
>>> Hoping you can give me some insight here.
>>>
>>> On windows when trying to import I get an exception. The application
>>> is trying to dereference a null pointer with the usual consequence.
>>>
>>> The problem is after loading the table object with the imported file
>>> it then tries to reference the cell location [-1,0] this returns a
>>> null pointer and thus crashes.
>>>
>>> I have done a trace through the code and have discovered the following.
>>>
>>> The csvdialog object has a field called  m_startLine which by default
>>> is set to 1.
>>> 101: m_startLine = 1;
>>>
>>> At line 640 in csvdialog.cpp it is set to default to zero if it
>>> doesn't exist in the profile. This seems ok and I assume the +1 covers
>>> a boundary condition
>>> 640: m_startLine = profilesGroup.readEntry("StartLine", -1) + 1; (was
>>> the -1 meant to be 0?)
>>>
>>> At line 908 it is used as a parameter and decremented to -1
>>> 908: redrawWindow(m_startLine - 1);
>>>
>>> At line line 1517 it is used to initialise 'm_startLine' which in turn
>>> is used to initialise 'row' at line 1546
>>>
>>> At line 1561, row (set to -1) is used to index the table
>>> 1561: label.setText(ui->tableWidget->item(*row*, col)->text());
>>>
>>> At that point it throws an exception as ui->tableWidget->item(row,
>>> col) returns a null pointer (by design) and there is no checking of
>>> the return value nor checking of the index before it is used.
>>>
>>> I assume this is a bug. What I am not sure about is what it should
>>> have been set to to start with -1, 0 or 1? as I said the default
>>> before loading a profile is 1.
>>>
>>> What really confuses me is on kubuntu I don't get an error with the
>>> same import file (actually it crashed the first time, but didn't again
>>> after I recreated the profile).
>>>
>>> Can you please advise where the error seems to be?
>>>
>>> Chris
>>
>>
>
>