Get-Version.ps1 doesn't work in 2.2.13

[the-ress]

Get-Version.ps1 fails with:

Exception calling "HelpFindLibGit2NativeBinaries" with "1" argument(s):
"Unable to find native binaries under directory: "[...]\NerdBank.GitVersioning\tools\..\build\MSBuildFull\lib\win32\x64"."

It seems 2.2.13 moved the lib directory (0da2e40), but the script wasn't updated.

[AArnott]

Thanks for the report. I haven't been testing the .ps1 scripts lately now that we have the nbgv dotnet CLI tool. Does that work for you?

[the-ress]

The .ps1 script returns an object. With nbgv I would have to parse its output.

[AArnott]

True. That does make the .ps1 a bit more convenient. However you can use nbgv for that purpose as well:

nbgv get-version -f json | ConvertFrom-Json

And voila! You have a powershell object again. :)

[the-ress]

Thank you, I didn't know get-version supported that.

[heaths]

The problem is that nbgv get-version returns the wrong information. For example, it shows the build version as 3.2.0.37 while get-version.ps1 correctly shows 3.2.37. Did defaults in code change that don't necessarily agree?

My version.json file:

{
  "$schema": "https://raw.githubusercontent.com/AArnott/Nerdbank.GitVersioning/master/src/NerdBank.GitVersioning/version.schema.json",
  "version": "3.3.0",
  "publicReleaseRefSpec": [
    "^refs/heads/master$"
  ],
  "cloudBuild": {
    "buildNumber": {
      "enabled": true,
      "includeCommitId": {
        "when": "nonPublicReleaseOnly",
        "where": "buildMetadata"
      }
    },
    "setVersionVariables": true
  }
}

[AArnott]

That's interesting. the ps1 script and the nbgv tool use exactly the same VersionOracle class to get the answers. I wonder if it's just a formatting thing.
FWIW, the version property in your version.json file looks misleading: it has a 3rd integer set to zero, but I think NB.GV ignores that. But it might be enough to throw off the nbgv tool and if so, is probably why I didn't catch it, since I'm not testing with an inert trailing .0 in my version property. @heaths if you wouldn't mind sharing your repo info with me, I can try it out.

[heaths]

I tried removing the ".0" and get the same result. See the develop branch of https://github.com/heaths/psmsi repo. Happens on an internal repo as well.

[AArnott]

I guess the trailing .0 isn't ignored any more anyway. NB.GV used to ignore that, but now it just moves height integer to the 4th position. Sorry I forgot about that.

Anyway, I can repro the .ps1 script failure, and also some inconsistencies in reporting versions by the various build/script/tools. So I'm working on that now.

[AArnott]

@heaths I tried commit a4b68111a on your repo (since your develop branch moved on since your report, evidently) and both nbgv and get-version.ps1 (after #258 and #256) report your version as 3.2.0.37, which given your version.json contains "3.2.0" appears to be the corrrect version.

For example, it shows the build version as 3.2.0.37 while get-version.ps1 correctly shows 3.2.37.

Why do you say 3.2.37 is the correct value when your version.json calls for "3.2.0" as the first three integers?

[heaths]

I did change it to "3.2", but, IIRC, may not have committed it, which would be required to make a difference in this case, right?

[AArnott]

Yes, in some cases at least, committing the version.json file is required.