Update DEVGUIDE to mention net472

[cartermp]

As per #6226 this is required

[cartermp]

This flaky test again:

Failed   Test request for parse and check doesn't check whole project
Error Message:
   Expected: true
Actual: false
  Expected: True
  But was:  False

Stack Trace:
   at FsUnit.shouldEqual[a](a x, a y) in D:\a\1\s\tests\service\FsUnit.fs:line 19
   at FSharp.Compiler.Service.Tests.PerfTests.Test request for parse and check doesn't check whole project() in D:\a\1\s\tests\service\PerfTests.fs:line 75

@dsyme can we consider disabling or removing this test? It's horribly unreliable and I don't know if it's worth having in the first place

[dsyme]

@dsyme can we consider disabling or removing this test? It's horribly unreliable and I don't know if it's worth having in the first place

Hmmmm I've not seen this test fail. But it's testing something pretty darn important (that we are not triggering entire rechecks of projects). We should at least add better diagnostics to work out what's failing and then dig into why

[<Test>]
let ``Test request for parse and check doesn't check whole project`` () = 

    let backgroundParseCount = ref 0 
    let backgroundCheckCount = ref 0 
    checker.FileChecked.Add (fun x -> incr backgroundCheckCount)
    checker.FileParsed.Add (fun x -> incr backgroundParseCount)

    let pB, tB = FSharpChecker.GlobalForegroundParseCountStatistic, FSharpChecker.GlobalForegroundTypeCheckCountStatistic
    let parseResults1 = checker.ParseFile(Project1.fileNames.[5], Project1.fileSources2.[5], Project1.parsingOptions)  |> Async.RunSynchronously
    let pC, tC = FSharpChecker.GlobalForegroundParseCountStatistic, FSharpChecker.GlobalForegroundTypeCheckCountStatistic
    (pC - pB) |> shouldEqual 1
    (tC - tB) |> shouldEqual 0
    backgroundParseCount.Value |> shouldEqual 0
    backgroundCheckCount.Value |> shouldEqual 0
    let checkResults1 = checker.CheckFileInProject(parseResults1, Project1.fileNames.[5], 0, Project1.fileSources2.[5], Project1.options)  |> Async.RunSynchronously
    let pD, tD = FSharpChecker.GlobalForegroundParseCountStatistic, FSharpChecker.GlobalForegroundTypeCheckCountStatistic
    backgroundParseCount.Value |> shouldEqual 5
    backgroundCheckCount.Value |> shouldEqual 5
    (pD - pC) |> shouldEqual 0
    (tD - tC) |> shouldEqual 1

    let checkResults2 = checker.CheckFileInProject(parseResults1, Project1.fileNames.[7], 0, Project1.fileSources2.[7], Project1.options)  |> Async.RunSynchronously
    let pE, tE = FSharpChecker.GlobalForegroundParseCountStatistic, FSharpChecker.GlobalForegroundTypeCheckCountStatistic
    (pE - pD) |> shouldEqual 0
    (tE - tD) |> shouldEqual 1
    (backgroundParseCount.Value  <= 9) |> shouldEqual true // but note, the project does not get reparsed
    (backgroundCheckCount.Value  <= 9) |> shouldEqual true // only two extra typechecks of files

    // A subsequent ParseAndCheck of identical source code doesn't do any more anything
    let checkResults2 = checker.ParseAndCheckFileInProject(Project1.fileNames.[7], 0, Project1.fileSources2.[7], Project1.options)  |> Async.RunSynchronously
    let pF, tF = FSharpChecker.GlobalForegroundParseCountStatistic, FSharpChecker.GlobalForegroundTypeCheckCountStatistic
    (pF - pE) |> shouldEqual 0  // note, no new parse of the file
    (tF - tE) |> shouldEqual 0  // note, no new typecheck of the file
    (backgroundParseCount.Value <= 9) |> shouldEqual true // but note, the project does not get reparsed
    (backgroundCheckCount.Value <= 9) |> shouldEqual true // only two extra typechecks of files
    ()

[cartermp]

I've seen it fail probably 10-20% of the time (see #4457 as well), but nothing in what it's testing gets changed in those PRs. So there's either a bug in FCS or it's a flaky test, but given recent history my first thought is it's the test.

[dsyme]

I've seen it fail probably 10-20% of the time (see #4457 as well), but nothing in what it's testing gets changed in those PRs. So there's either a bug in FCS or it's a flaky test, but given recent history my first thought is it's the test.

OK thanks. We should dig into what's failing by adding better diagnostics