The reason of the crash is a static_cast of an unknown type that causes undefined behaviour. The feature AdditiveLoft has the
property Section of type PropertyLinkSubList but the function does a static_cast to PropertyLinkList.
The solution is to use a dynamic_cast that returns null if the cast fails.
This patch substitutes by isAttachedToDocument() (almost) everywhere where
getNameInDocument() is used for this purpose.
The very few places not touched by this patch demand a (just a little) less trivial change.
When we change the returning type of getNameInDocument() to std::string,
those places will be easily found, because they shall generate a compiler error
(converting std::string to bool).
Rationale:
The fact that getNameInDocument() return nullptr to indicate
that the object is not attached to a document is responsible for lots of bugs
where the developer does not check for "nullptr".
The idea is to eliminate all those uses of getNameInDocument() and, in the near future,
make getNameInDocument() return always a valid std::string.
* [PD] Add new need active body dialog
Intended for use wherever an active body is needed but none exists. The dialog
contains a list with the bodies present in the document, with an extra option to
make a new body. Custom text can be provided if needed.
This commit also replaces the warnings used by "new sketch" and "primitive"
command with this new dialog.
Addresses issue #4288.
As reported by @donovaly, if you try to move a pipe that does not have
an auxilliary spine, the code segfaults. This commit adds a check to ensure
that not only does the property exist (which in this case it always does), but
also that it does not contain a null. That is a valid value for the property
to have, but cannot be itself interrogated for moveability by the isFeatureMovable
recursive call. This fix is also applied to the other similar conditions in that
function to ensure they never yield the same segmentation fault.
Forums topic: https://forum.freecadweb.org/viewtopic.php?p=479388#p479388
With the recently introduced AttachExtension the PartDesign primitives can be attachable by themself and don't need anymore the helping coordinate system datum. This simplyfies the code and the TreeView largely
AttachableObjects are desired in multiple occasions, and the current AttachableObject is not flexible enough to handle all cases. Hence the code is portet to an extension, which gives the needed flexibility.
Due to problems onthe windows platform the virtual inheritance approach must be dropped. NExt to the already reimplemented proeprty interface the Type interface is reimplemented too. This change allows to revert some earlier changes.
