QGIS » QGIS Application

Sorry, there's a mistake in the script.

Instead of

newAttributes = []

it should be

newAttributes = {}

The crash is probably due to incorrect format of newAttributes. The values should be QVariant instances. (But it should not crash either.) What if you use:

newAttributes = { paramindex : QVariant("newparam"), valueindex : QVariant("newvalue") }

As a side note (not related to the crash), you don't need to use startEditing() and commitChanges() if you are modifying the provider directly (i.e. not using vector layer's edit buffer).

Also, you should always call provider's select() function before any actual nextFeature() calls. The rewind() method is useless here, too.

Giuseppe:

Yes, my original script crashes master too. I've just installed the Windows build 70273b9 to test it.

Martin:

First of all, thank you very much for your comments. It clarifies some doubts I had about startEditing(), rewind() or select()

I changed the script to use Variant instances and now the update works OK.

Thanks again to you and Giuseppe for your support.

Martin Dobias wrote:

The crash is probably due to incorrect format of newAttributes. The values should be QVariant instances. (But it should not crash either.)

I've got crashes using the snipped above, the crash occurs in convertions.sip file.

Martin, I've seen that you didn't used sipConvertToMappedType for QMap<int, TYPE> since you got crashes (it was 2008), you replaced it with a snipped that manually converts a PyObject to QMap<int, TYPE>:

// using sipConvertToMappedType to convert directly to QMap<int, TYPE> doesn't work
// and ends with a segfault

In my Ubuntu 12.10, sip 4.13.3, using sipConvertToMappedType it works without any crash, probably it was fixed in somewhat sip version.
A #ifdef SIP_VERSION >= 0x041200 could help, we already know we get crashed using the actual code in SIP >= 4.12 (as it crashes on Win as well).

Martin, I guess there's more than one problem in the current conversions.sip file...

Have a loot at lines 259-269:

    //TYPE *t = reinterpret_cast<TYPE *>(sipConvertToInstance(PyList_GET_ITEM(sipPy, i), sipClass_TYPE, sipTransferObj, SIP_NOT_NONE, &state, sipIsErr));
    QList<TYPE> * t = reinterpret_cast< QList<TYPE> * >(sipConvertToMappedType(PyList_GET_ITEM(sipPy, i), qlist_type, sipTransferObj, SIP_NOT_NONE, &state, sipIsErr));

    if (*sipIsErr)
    {
      sipReleaseInstance(t, sipClass_TYPE, state);
      delete ql;
      return 0;
    }
    ql->append(*t);
    sipReleaseInstance(t, sipClass_TYPE, state);

The TYPE* t variable was commented and replaced with QList<TYPE> * t, but the sipReleaseInstance is assuming to work on the old t variable...

I don't know if it makes sense to fix that and other similar problems or waiting for your merge...

Giuseppe Sucameli wrote:

I've got crashes using the snipped above, the crash occurs in convertions.sip file.

FYI, it crashes at line 627,

sipReleaseInstance(tobj2, sipClass_TYPE, state);

I'm not a SIP guru, but probably the problem is the same I reported before, since tobj2 should be fa I suppose...