Natvis support

[GloriousPtr]

Is there a plan to add natvis support?

[ryanfleury]

We'll likely not support all NatVis features, but our working idea is: In the PDB -> RADDBG conversion process we'll parse the NatVis and convert it down to a much simpler & more constrained format (not supporting some of the more insane NatVis features, but supporting most of the common ones), which can then be used as "auto-filled view rules" for specific types. This should cover many of the common cases for NatVis, and keep the debugger's frontend on only one visualization engine.

[mnicolella]

I'm not sure which things you mean by "some of the more insane NatVis features" but often there are opaque handle types, or indices, or 'keys' which require complex logic to visualize what they refer to - sometimes things like a hash table lookup, etc...

Preferably, it would be great if we could ask the debugger to execute a function in the process in order to do the lookup to get the thing that should be visualized. But the VS debugger generally doesn't allow that from natvis. Alternatively, natvis supports some complicated stuff that allows these visualizations to be written.

Personally I would be fine with a complete alternate to natvis but I understand the desire to make it compatible with existing stuff.

[FaultyPine]

Seconding this. I'd guess a pretty large amount of people aren't using this because of their deeply important natvis debugging setup. Having this at least reasonably supported would be huge for transitioning people away from VS debugger.