[cpyrt] Drop the dead ROOT signal-longjmp guard

[aaronj0]

No description provided.

[vgvassilev]

We should not drop that but implement it properly.

[guitargeek]

We should not drop that but implement it properly.

I would not implement things just for the fun of it: if this was dead code before, it was not relevant to users so this is a strong signal that we should not implement it anymore, and only come back to this if there is an explicit user request.

[guitargeek]

LGTM!

Consider opening an issue about crash isolation for C++ calls, potentially with out-of-process execution, as a possible feature.

In any case, the removal of the dead signal-longjmp guard is the first step that should be merged.

[vgvassilev]

We should not drop that but implement it properly.

I would not implement things just for the fun of it: if this was dead code before, it was not relevant to users so this is a strong signal that we should not implement it anymore, and only come back to this if there is an explicit user request.

IIUC, this is dead code after the rework and not dead in cppyy. That means that it was used in standalone cppyy for recover from crashes.

[aaronj0]

I would not implement things just for the fun of it: if this was dead code before, it was not relevant to users so this is a strong signal that we should not implement it anymore, and only come back to this if there is an explicit user request.

IIUC, this is dead code after the rework and not dead in cppyy. That means that it was used in standalone cppyy for recover from crashes.

Yes in cppyy proper, CPyCppyy was supplied the jump target from cppyy-cling. This never worked since the start of the compiler-research forks so I meant dead code in that regard. I've taken a stab at having CppInterOp provide this for cppjit in the following branches:

https://github.com/aaronj0/CppInterOp/tree/signal-guard
https://github.com/aaronj0/cppjit-compres/tree/signal-guard

but they are a WIP. I can open an issue tracking this.