QGIS » QGIS Application

It does not include any CSV files. I'm not sure if I can attach the project/data as the size reaches 92MB. QGIS only crashes when I stress it. If I navigate the canvas in a moderate fashion, it works very well.

Per Gammerath wrote:

It does not include any CSV files. I'm not sure if I can attach the project/data as the size reaches 92MB. QGIS only crashes when I stress it. If I navigate the canvas in a moderate fashion, it works very well.

a project and data (even sent privately) would be very useful to try replicate the issue and get the error message that can lead to the source of the problem.

I just tried with the Alaska sample dataset, and I could easily make it crash too by zooming violently. So I'm quite sure it's about the data size and not the layer properties or so.

Per Gammerath wrote:

I just tried with the Alaska sample dataset, and I could easily make it crash too by zooming violently. So I'm quite sure it's about the data size and not the layer properties or so.

can you please attach a project based on the alaska sample dataset (relative paths)? thanks.

I didn't save the project. I just started a new project and added all the Alaska shapes. Like I said, I'm sure it will crash with any project that contains large datasets. I tried loading one shapefile at a time, and the more shapefiles I loaded, the more likely QGIS would crash.

I'm having a similar issue on Fedora 20 x64, running QGIS 2.4.0. The project I'm working with does have csv files with WKT geometry, both points and lines.

When I run qgis from the console, zooming around quickly results in the message:

QGIS died on signal 11

(which I believe indicates a segfault)
And then QGIS is totally unresponsive until I kill it with signal 15. At one point it also froze my entire system.

Are issues like this serious enough to merit a point release?

Matthew Bush wrote:

I'm having a similar issue on Fedora 20 x64, running QGIS 2.4.0. The project I'm working with does have csv files with WKT geometry, both points and lines.

see #10744

what about if you remove the CSV files?

If I remove the CSV files, then there's nothing left except google maps from the openlayers plugin.

I created a new project and zoomed around trying to break it. When I had several layers, each either in shp format or coming from the openlayers plugin, everything worked fine. One of the shp layers contained point geometry information. I added a CSV file with WKT line geometry information, and everything still worked. Once I added another CSV file with WKT point geometry information, I was able to get qgis to crash (hangs on signal 11) in less than a minute.

Matthew Bush wrote:

If I remove the CSV files, then there's nothing left except google maps from the openlayers plugin.

I created a new project and zoomed around trying to break it. When I had several layers, each either in shp format or coming from the openlayers plugin, everything worked fine. One of the shp layers contained point geometry information. I added a CSV file with WKT line geometry information, and everything still worked. Once I added another CSV file with WKT point geometry information, I was able to get qgis to crash (hangs on signal 11) in less than a minute.

so your case seems to fit #10744

Giovanni Manghi wrote:

so your case seems to fit #10744

Mostly. It may still be related to the file size of the project, as the CSV layers I was working with were significantly larger than the SHP layers.

This is a large enough bug to prevent my organization from upgrading to 2.4, despite the huge benefit of multithreaded rendering. Assuming that it does get fixed, how is said fix likely to be released? As a point release? Or should we plan on using a 2.5 daily?

Have you tried to reduce the number of max core to use in multithreaded rendering (rendering tab on option window)? I have experienced frequent hangs (no CSV layer involved) using multicore rendering, but the problem seems to be resolved using only 2 core for multithreaded rendering.

Alessandro Ciali wrote:

Have you tried to reduce the number of max core to use in multithreaded rendering (rendering tab on option window)? I have experienced frequent hangs (no CSV layer involved) using multicore rendering, but the problem seems to be resolved using only 2 core for multithreaded rendering.

I can confirm, that QGIS definitely becomes stable when rendering is limited to 2 cores.

Matthew Bush wrote:

Mostly. It may still be related to the file size of the project, as the CSV layers I was working with were significantly larger than the SHP layers.

do you still get crashes if disabling the "renders layers in parallel using many CPS cores" option or at least choosing a low number of cores to use?

This is a large enough bug to prevent my organization from upgrading to 2.4, despite the huge benefit of multithreaded rendering. Assuming that it does get fixed, how is said fix likely to be released? As a point release? Or should we plan on using a 2.5 daily?

I guess (from what I have read in threads in the dev mailing list) that is also a matter of resources, if someone is willingly to invest then dot releases/backports may be possible.

You mention openlayers plugin, if you remove the openlayers plugin and use max# or cores does the crash happen? The openlayers plugin does not work well with qgis 2.4, this is a frequently reported problem.

Etienne Tourigny wrote:

You mention openlayers plugin, if you remove the openlayers plugin and use max# or cores does the crash happen? The openlayers plugin does not work well with qgis 2.4, this is a frequently reported problem.

there are anyway also crashes without OL layers and with MTR on, like in #10744

I still get segfaults with max_cores=2, when I have multiple CSV layers, one of which contains point geometry. It worked fine with three layers, all lines (size 321MB, 12MB, 686KB), but once I added the fourth layer, with points, (size 63MB) it crashed.

This time I created the project from scratch in 2.4, and didn't use openlayers.

edit: added file sizes

please try again on qgis master, the problem about CSV files has been solved and it may have solved also the one reported here.

I've compiled the release_2_4 branch from source, and this is resolved for me.

When is this likely to make it into the binary installers?

can the original reporter leave feedback if a test on qgis master has been done? thanks!