Incompatible plugins and seg faults on start. Was: Re: [Qgis-developer] Seg fault on program start (r10258)

[Martin Dobias]

2009/3/4 Borys Jurgiel <borysiasty at aster.pl>:
> On środa 04 marzec 2009 08:13:48 Craig Leat wrote:
>> Thanks for the advice. It works now :-)
>
> Btw. An incompatible plugin shouldn't cause program crashes! After upgrading
> from the 1.0 to the 1.1 it will probably affect all users who have GridMaker
> enabled. Also external plugins can cause troubles after QGIS upgrade.
>
> Shouldn't we implement any test for plugin version? I'm not sure what is the
> right way. A revision number of QGIS the plugin was compiled against? Probably
> there is something much smarter.

I think we should add another function to C++ plugins, something like
"requiredQgisVersion" and this one would return the version of QGIS it
has been compiled with (from QGIS_VERSION macro). Then the first thing
when loading a plugin would be to resolve this function and compare
the actual qgis version and plugin's qgis version.

Otherwise we might be getting more complaints from users. This would
solve also a longer standing problem that if there are some plugins
for older version of qgis (and there are also libs for the older
version), QGIS tries to load them anyway and some incompatibility
causes python bindings not to be loaded (because QgsApplication
returns incorrect paths).

Martin