Debug points failing to bind in body of "[ for x in xs -> body ]"

[dsyme]

See #12432 (comment)

It's a good question - do you have a more specific repro of this problem please?

Just tested with the following demo:

[<EntryPoint>]
let main args =

    let test1 () =
        [ 3; 2; 1 ]
        |> List.map (fun x -> x + 10)
        |> List.sort

    let test2 = [
        for x in test1() do
            let xx = x * x
            printf "test"
            xx
        ]

    let test3 = [
        for x in test1() ->
            let xx = x * x
            printf "test"
            xx
        ]

    printf "The end"

Notice that:

• I can add breakpoints in both test2 and test3 before starting the debugger (F5)
• When debugging, only the breakpoints in test2 stop the debugger
• If I remove/re-add a breakpoint, it works in test2, it fails with the message "The breakpoint failed to bind." in test3

I'm using Visual Studio Community 2022 (64-bit) - Version 17.1.0, and it's a .NET 6.0 project.

[ntovas]

Can't reproduce in VS: Community/Enterprise 2022 (64-bit) - Version: 17.4.0 preview 2.1/latest main branch with either .net 6.0/.net 7.0 rc1

Project Type: Console
Code:

[<EntryPoint>]
let main args =

    let test1 () =
        [ 3; 2; 1 ]
        |> List.map (fun x -> x + 10)
        |> List.sort

    let test2 = [
        for x in test1() do
            let xx = x * x
            printf "test"
            xx
        ]

    let test3 = [
        for x in test1() ->
            let xx = x * x
            printf "test"
            xx
        ]

    0

[brianrourkeboll]

I ran into this and other debug point oddities in collection expressions in #17067 (comment) — in fact, although the repro code above is now optimized to a fast while-loop, I kept the debug points exactly the same as before, which means that this still reproduces.

(It also reproduces if you write for x in Seq.ofList (test1()) do so that the optimization doesn't apply.)

It would be nice if there were a way to write tests for debug point locations. If that is already done somewhere, I haven't seen it.

[kirisky]

The bug is still there. Is anyone working on this? If not, I'd like to give it a try.

[brianrourkeboll]

The bug is still there. Is anyone working on this? If not, I'd like to give it a try.

I don't think anyone is looking at this currently, no.

Ideally we would:

1. Document how debugging should behave in each scenario (for-to, for-in over ranges, lists, arrays, sequences).
2. Write tests asserting that behavior.
3. Update the codegen where needed to comply with that behavior.

You can see some examples of the differing behaviors in #17067 (comment).

[kirisky]

Thanks, Brian! Will take a look!