[mythtv] [mythtv-users] Android mythfrontend mysql driver problems

[Mark Spieth]

On 10/11/2022 9:14 am, Mark Spieth wrote:
> On 10/11/2022 3:10 am, David Engel wrote:
>> On Wed, Nov 09, 2022 at 06:23:54PM +1100, Mark Spieth wrote:
>>> On 3/11/2022 9:57 am, David Engel wrote:
>>>> On Thu, Sep 29, 2022 at 03:01:25PM +1000, Mark Spieth wrote:
>>>>> I updated mariadb connector to 3.2.6 but I havent seen any issues 
>>>>> with
>>>>> recent builds. Last one I did was end of May using the previous 
>>>>> build env
>>>>> settings if SDK 28, NDK 21.4... and tools 29.0.3. I remember a 32 
>>>>> bit build
>>>>> issue I was fixing at the time.
>>>> Mark,
>>>>
>>>> I tracked down a problem I saw to mariadb connector 3.2.6. Using the
>>>> default, nvshield config, my resulting apk fails silently when
>>>> deleting recordings.  By that, I mean the recording gets removed from
>>>> the list in playbackbox but as soon as there's a refresh or
>>>> playbackbox is re-entered, the recording is back.
>>>>
>>>> ENOTIME kept prventing from working on the problem a lot but I finally
>>>> tracked it down to the query in ProgramInfo::ProgramInfo(uint
>>>> _recordedid).  Specifically, the starttime that gets read from the
>>>> database is empty/blank.  That leads to it being set to the current
>>>> time in PI::clear().  Ultimately, the backend is not able to look up
>>>> the correct recording when it gets the delete request.
>>>>
>>>> The problem is fixed for me by switching back to mariadb connector
>>>> 2.3.7.
>>>>
>>> I have the qt date fix in my myth master tree and I havent seen the 
>>> problem
>>> since I added that. Before that, when I deleted a program, went back 
>>> to main
>>> menu and then back into playbox the program was still there.
>>>
>>> Probably the same problem (timezone info in rendered datetime string)
>> Good you could reproduce and fix it.  When do you think you'll commit
>> it?
>
> There is a PR and issue already in github for a long time. Its a myth 
> fix as it affects mysql as well.
>
> https://github.com/MythTV/mythtv/pull/574
>
> Its in its own branch feature/datetime-sql-bind-format-fix
>
> I'm hesitant to merge myself unless there is a consensus. My opinion 
> is that its easier to fix in myth and then its fixed everywhere.

Update on the delete issue.

I have just updated my full system(s) this weekend and can now see 
delete failing again. reverting back to an android build from 6 oct also 
fails to delete, and this one worked before. So the issue must live in 
the backend. The linux frontend deletes fine and so does leanfront. so 
there is some sort of compatability issue somewhere.

The version of the backend that was working was built on 31 may 2022 so 
that leaves a lot of search space.

This is with the above patch and I also tried the qt patch for the qt 
mysql driver, with the same behaviour.

I will look into this unless someone else has some but like david, ENOTIME.

cheers

Mark