[lldb-dap] REPL auto-detection does not recognize scripted command containers

[vogelsgesang]

If I add a custom command container via

command container add -h 'Debugging utilities for C++20 coroutines' coro
command script add -o -p -c {__name__}.CoroBacktraceCommand coro bt

the auto-detection of the REPL mode for coro bt does not work as expected.
lldb-dap thinks this would be an expression, although it should be treated as a LLDB command.

For a full debugger script, see the example script in #142651 (inside the DebuggingCoroutines.rst)

[llvmbot]

@llvm/issue-subscribers-lldb-dap

Author: Adrian Vogelsgesang (vogelsgesang)

If I add a custom command container via

command container add -h 'Debugging utilities for C++20 coroutines' coro
command script add -o -p -c {__name__}.CoroBacktraceCommand coro bt

the auto-detection of the REPL mode for coro bt does not work as expected.
lldb-dap thinks this would be an expression, although it should be treated as a LLDB command.

For a full debugger script, see the example script in #142651 (inside the DebuggingCoroutines.rst)

[ashgti]

This is due to command script add -o -p -c {__name__}.CoroBacktraceCommand coro bt not having the -C or --completion-type flag.

I think it should work if you add that flag appropriately for your script.

[vogelsgesang]

I tried using command script add -o -p -C variable-path -c {__name__}.CoroBacktraceCommand coro bt instead. That gives an error "invalid combination of options for the given command".

Also, afaict, the -C parameter would only influence the auto-completion of parameters passed to coro bt. Why should it influence whether lldb-dap detects the presence of the coro command in DetectReplMode?

[vogelsgesang]

🤔 DetectReplMode calls UserCommandExists, but it doesn't call UserMultiwordCommandExists... Should it?

CC @labath as the author of 33bf08e which introduced the usage of UserCommandExists

[ashgti]

Ya, sorry, your right it looks like this is going through a code path that isn't checking for lldb_private::CommandIntepreter::m_user_mw_dict map for command contains.

[labath]

🤔 DetectReplMode calls UserCommandExists, but it doesn't call UserMultiwordCommandExists... Should it?

I guess it should -- but it looks like that one is not plumbed up to the SB API. Instead of plumbing that, it may be better to create a single "does this string represent a valid command (of any kind)" function...