"readOnlyTarget" capability

[roblourens]

Nuclide has a "readOnlyTarget" capability flag which disables some unsupported features of the debugger, for debugging static dumps. This should exist in the debug protocol too.

Instead of a master flag like "readOnlyTarget", we would prefer more specific flags for individual features. But capabilities usually enable features, not disable them, and it seems weird to have a flag like supportsStepping which must be set 99% of the time.

@mostafaeweda
@ebluestein

Which features are disabled by this flag? Do you think there would be different features supported in different situations (so they'd need to be disabled on a more granular level)?

VS docs: https://msdn.microsoft.com/en-us/library/d5zhxt22.aspx

[ebluestein]

This flag disables things like next/stepIn/stepOut/continue, as well as set variable. I don't think we currently disable evaluate, but that might depend on the back-end if it can evaluate in a static dump.

My primary use case for this is debugging a core dump with GDB / LLDB - but other debuggers could use this as well. For example, it's possible to attach in read only mode to a desktop Java app that wasn't started with a debug agent listening (in that mode Java lets you break in and see threads and call frames, but not modify anything or resume - it's basically for debugging deadlocks)

Maybe LaunchResponse / AttachResponse could have an optional isReadOnly boolean in its body? I agree capabilities feels like a strange place for this