[Qgis-developer] Problem with the MySQL/GDAL implementation in QGIS

[Pavol Hnila]

Hi,

After having tested builds based on particular commits of QGIS source
code I can confirm that Matthias is right. It must be the commit below
(or something related to it implemented on the same day) that
instructs OGR to open and close the dataset for every record and thus
causes substantial delays.

The build of 0b9d1dc (from 21st June 2013) results in a QGIS crash
with segfault. Yet the preceding commit for qgsogrprovider.cpp
(9bbc647 from 20th June 2013) still works perfectly, while the
successive commit (6876540 from 21st June 2013) has already the
described OGR excessive reopenings problem.

Is there a chance that someone corrects this particular code for the
next release?

Best wishes,
Pavol

P.S.: According to what I know the problem is reproducible with any
MySQL or PostGIS table connected via OGR. Should someone need a MySQL
testing environment for this purpose, contact me please off-list.


On 11.05.2015 09:41, Matthias Kuhn wrote:
> 
> 
> On 05/11/2015 09:29 AM, Pavol Hnila wrote:
>> The principal problem pointed out by Even however remains - the 
>> repeated opening and closing of the OGR datasource for every
>> record in the table. I will try to figure out which nightly build
>> between 1.8 and 2.0 release caused this change in behaviour.
>> Since I cannot figure it out in the code and need to try
>> manually, this can take some time. Unless someone else on the
>> list can figure it out from the master code ...
> It was most likely that one: 
> 0b9d1dc49b3277a842d0a18a2f1ba99782965999
> 
> Matthias
>