Re: Heisenbug-ish NullPointerExceptions in DataContext.flushToParent() [solved]

[Hugi Thordarson <hu...@karlmenn.is>]

To provide some closure on this issue, in case anyone looks at this thread in the future:

Turns out Cayenne was totally innocent and this was just a symptom of a simple "Internet double-clicker" issue.

A button in my web application invoked a method that deletes an object. If the user double-clicked that button -> the same method was invoked twice -> the later invocation tried to do things with a deleted object -> it's snapshot had been evicted from the cache -> kaboom!

But thanks still for the help John! :)

Cheers,
- hugi



> On 7 May 2021, at 20:21, Hugi Thordarson <hu...@karlmenn.is> wrote:
> 
> Rethinking this a bit… If it turns out that the snapshot has actually been evacuated from cache, how should I obtain an instance of the snapshot? And isn't it a bit odd that a snapshot has been evacuated for an object that's about to be updated or is that to be expected?
> 
> Cheers,
> - hugi
> 
> 
> 
>> On 7 May 2021, at 19:55, Hugi Thordarson <hu...@karlmenn.is> wrote:
>> 
>>> On 7 May 2021, at 13:34, John Huss <jo...@gmail.com> wrote:
>>> 
>>> On Fri, May 7, 2021 at 8:18 AM Hugi Thordarson <hu...@karlmenn.is> wrote:
>>> 
>>>> Well I'll be… Over two decades of full time Java and I somehow didn't know
>>>> about that stack trace omission. Yikes. Thanks!
>>> 
>>> You’ve just never caused an error that bad before right? 😉
>> 
>> *cough* yes. Of course, that must be it 😁
>> 
>>>> ....
>>>> So it seems either 'snapshot' or 'objAttribute' is null, probably the
>>>> snapshot, right? Any ideas why that might be?
>>>> 
>>> 
>>> Yeah, it would be the snapshot is null. The snapshot cache is limited in
>>> size, so it may have purged that entry if you have a busy server. You could
>>> try increasing the size of the snapshot cache:
>>> 
>>> org.apache.cayenne.configuration.Constants.SNAPSHOT_CACHE_SIZE_PROPERTY
>> 
>> Good thinking, but I've already got a ginormous snapshot cache (3.000.000) for the traffic, so I don't think that's it.
>> Going to hammer away on this code path a bit and see if I can replicate the error.
>> 
>> 
>>>> 
>>>> - hugi
>>>> 
>>>> 
>>>> 
>>>>> On 7 May 2021, at 12:53, John Huss <jo...@gmail.com> wrote:
>>>>> 
>>>>> Or if you have older logs, you can look back farther to find one without
>>>>> the cause omitted.
>>>>> 
>>>>> On Fri, May 7, 2021 at 7:52 AM John Huss <jo...@gmail.com> wrote:
>>>>> 
>>>>>> It will omit the trace if it has happened a very large number of times.
>>>> If
>>>>>> you restart the server it will start printing it again when it happens.
>>>>>> 
>>>>>> On Fri, May 7, 2021 at 7:26 AM Hugi Thordarson <hu...@karlmenn.is>
>>>> wrote:
>>>>>> 
>>>>>>> One would think so, but there's just no stack trace printed for the
>>>>>>> NullPointerException.
>>>>>>> 
>>>>>>> 
>>>>>>>> On 7 May 2021, at 12:24, John Huss <jo...@gmail.com> wrote:
>>>>>>>> 
>>>>>>>> The first line of the stack after the Caused By line would tell you
>>>> the
>>>>>>>> line with the problem, wouldn't it? What was that line?
>>>>>>>> 
>>>>>>>> On Fri, May 7, 2021 at 7:16 AM Hugi Thordarson <hu...@karlmenn.is>
>>>>>>> wrote:
>>>>>>>> 
>>>>>>>>> Hi all,
>>>>>>>>> 
>>>>>>>>> occasionally my users encounter errors when committing changes to the
>>>>>>> DB.
>>>>>>>>> The relevant part of the stack trace looks like this:
>>>>>>>>> 
>>>>>>>>> ---------
>>>>>>>>> org.apache.cayenne.CayenneRuntimeException: [v.4.1 Jul 14 2020
>>>>>>> 10:26:08]
>>>>>>>>> Commit Exception
>>>>>>>>>    at
>>>>>>>>> 
>>>>>>> 
>>>> org.apache.cayenne.access.DataContext.flushToParent(DataContext.java:774)
>>>>>>>>>    at
>>>>>>>>> 
>>>>>>> 
>>>> org.apache.cayenne.access.DataContext.commitChanges(DataContext.java:691)
>>>>>>>>>    [...removed rest of trace…]
>>>>>>>>> Caused by: java.lang.NullPointerException
>>>>>>>>> ---------
>>>>>>>>> 
>>>>>>>>> Unfortunately I've been unable to reproduce this error, so before I
>>>>>>> dive
>>>>>>>>> into it I'm wondering if this looks familiar to anyone, if there are
>>>>>>>>> obvious places to look or if there are recommendations for strategies
>>>>>>> to
>>>>>>>>> debug?
>>>>>>>>> 
>>>>>>>>> Cheers,
>>>>>>>>> - hugi
>>>>>>> 
>>>>>>> 
>>>> 
>>>> 
>> 
>