I have tried the debugger in VS Code with some simple scripts and it seems to work just fine. However, when I try to debug some production code that uses “large” packages (DifferentialEquations, Plots and a few more), it seems like the debugger hangs. I suspect that in reality it is just VERY busy interpreting (as opposed to compiling) these large libraries, but I don’t know enough of the inner workings to be certain.
I am trying to find a subtle bug in a set of differential equations for a reactor model that has very non-trivial (as in several pages of code) kinetics, so a debugger would be a blessing here.
If anyone has some experience in using the debugger in code that uses these (or similar) libraries and cares to share some tips and best practices, I would love to hear from you.
If you don’t need breakpoints, use the
Compiled Mode toggle in the breakpoints section:
If you do, consider putting them before expensive operations and then stepping to the target location.
Now, if that’s also not possible, consider giving Infiltrator.jl a go, which drops you into a REPL session at your breakpoint but doesn’t allow any further stepping.
Naive question but what’s the typical debugging workflow with the debugger but without breakpoints? (I can imagine lots of ways to debug in general, but I must be missing something obvious because the obvious use of a debugger seems to be set breakpoint => run to breakpoint => step through code in debugger.)
Skip the first two steps?
Mostly useful only when you can start debugging close to where you want (or just extract the part that you’re interested in). But yeah, obviously that’s a big limitation and hopefully we’ll get some big improvements in the future (e.g. combining Infiltrator.jl and Debugger.jl).
If a breakpoint is made after a time consuming segment of code, it is much slower than stepping through to that point?
So it is faster just to do a @enter and move down to your desired point? (than using break points)
So the only distinction in runtime is whether you’re running in compiled mode or not. If you are, breakpoints that are not in the current local scope won’t work.
Can you switch between compiled mode and not inside of one debugging session? (The
compiled mode check box seems to be checkable, but it’s not obvious when the results take effect: immediately? Next
Juno.@enter? On Julia restart?
In compiled mode, does stepping to a selected line work, and would that function much like a breakpoint? (Albeit not a conditional breakpoint)?
Yeah, that takes effect immediately.
In that case a breakpoint should actually work fine, provided you don’t pass any function boundaries. But otherwise just hit “Step Over” a few times and you should be good to go.