gno
gno copied to clipboard
feat: add an interactive debugger to GnoVM
We provide here an embedded interactive debugger to let the user control and inspect its program at symbolic level, with the same features and commands as classical debuggers: gdb, lldb or delve.
The debugger is enabled by setting the -debug
flag in gno run
command, which loads the target program and immediately shows a debugger prompt on the console:
$ gno run -debug /tmp/my-program.gno
Welcome to the Gnovm debugger. Type 'help' for list of commands.
dbg>
Providing -debug-addr
flag allows to start a remote debugging session, and not interfer with the program stdin and stdout. For example, in a first terminal:
$ gno run -debug-addr :4000 /tmp/my-program.gno
Waiting for debugger client to connect at :4000
And in a second terminal, using a netcat like nc(1):
$ nc localhost 4000
Welcome to the Gnovm debugger. Type 'help' for list of commands.
dbg>
The debugger works by intercepting each execution step at virtual machine level (each iteration within Machine.Run
loop) to a callback, which in turns can provide a debugger command REPL or check if the execution can proceed to the next step, etc.
The general logic and structure is there. It is possible to continue
, stepi
, detach
, print
, stack
, etc and get a general feedback of the user experience and the impact on the code.
Efforts are made to make this feature minimally intrusive in the actual VM, and not interfering when the debugger is not used.
It is planned shortly after this PR is integrated to add the capacity to attach to an already running program, and to taylor the data format for existing debugging environments such as VScode, etc, as demand arises.
Resolves gnolang/hackerspace#54
Contributors' checklist...
- [x] Added new tests, or not needed, or not feasible
- [x] Provided an example (e.g. screenshot) to aid review or the PR is self-explanatory
- [x] Updated the official documentation or not needed
- [x] No breaking changes were made, or a
BREAKING CHANGE: xxx
message was included in the description - [x] Added references to related issues and PRs
- [x] Provided any useful hints for running manual tests
- [ ] Added new benchmarks to generated graphs, if any. More info here.
I have found some issues when computing the stack from a non func block (like if
or for
body).
I also found some issues when printing a global var.
I would like also to set a test suite for the various debugger commands and in the various possible contexts such as the ones mentioned above. That requires a bit more work, maybe in a new PR (still pondering).
Codecov Report
All modified and coverable lines are covered by tests :white_check_mark:
Project coverage is 48.43%. Comparing base (
977a3f4
) to head (50c3edf
).
Additional details and impacted files
@@ Coverage Diff @@
## master #1563 +/- ##
==========================================
- Coverage 55.01% 48.43% -6.58%
==========================================
Files 481 409 -72
Lines 67432 61940 -5492
==========================================
- Hits 37097 30001 -7096
- Misses 27318 29442 +2124
+ Partials 3017 2497 -520
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
Yes @thehowl, I will now look at unifying common parts as you suggest. Thanks for your thorough review and useful suggestions.