Hi, > I checked the change log for the updated 1.8.15-3 version of the package > and saw the "fix memory leak" entry earlier.
That was another memory leak. That's why this bug is not closed. > So I updated it, switched > back to qtcurves, ulimit-ed the virtual memory to 1.2GB, and started > mysql-workbench from that shell. > > After getting to the SQL editor, the virtual memory was at 1GB (compared > to 600MB for Adwaita), and after the 4th query it started it's explosive > memory leak, same as before. So it would appear there is still an issue > in the qtcurve libs, or maybe workbench abusing the theme in some > strange way. For kicks and giggles I ran 228 simple queries with the > Adwaita theme, and the virtual memory only went up by about 50MB, but I > suspect that's for the command and result history. > > There is the gtk_tree_view_unref_tree_helper `node != NULL` assertions > that shows up on the terminal after the second query and every query > thereafter. Perhaps there is still a memory leak from a local pointer > getting reset without freeing memory-intensive qtcurve objects first. Yes, it looks that bug can be related with processing GtkTreeView elements, but nor I nor upstream have no enough free time to research the problem right now. Hope we will find it soon. Best regards, Boris
