jpm deps behavior on Windows

[sogaiu]

With cf6086ae3b282046141fd7b6fbd1d08edcb85fdd on Windows, jpm deps for a project with at least one dependency can give output like this:

C:\Users\user\Desktop\jpm-sample>jpm deps
From https://github.com/janet-lang/spork
 * branch            master     -> FETCH_HEAD
Already up to date.
HEAD is now at 98f61de Merge pull request #44 from subsetpark/streams
C:/Users/user/scoop/apps/git/2.24.1.windows.2/mingw64/libexec/git-core\git-submodule: line 7: sed: command not found
C:/Users/user/scoop/apps/git/2.24.1.windows.2/mingw64/libexec/git-core\git-submodule: line 21: .: git-sh-setup: file not found
error: command failed with non-zero exit code 1
  in os/execute [src/core/os.c] on line 1049
  in shell [C:/Users/user/AppData/Local/Apps/Janet/Library/jpm/shutil.janet] (tailcall) on line 107, column 5
  in download-bundle [C:/Users/user/AppData/Local/Apps/Janet/Library/jpm/pm.janet] on line 167, column 10
  in bundle-install [C:/Users/user/AppData/Local/Apps/Janet/Library/jpm/pm.janet] on line 179, column 13
  in deps [C:/Users/user/AppData/Local/Apps/Janet/Library/jpm/commands.janet] (tailcall) on line 123, column 7
  in _thunk [C:\Users\user\AppData\Local\Apps\Janet\bin\/jpm] on line -1, column -1
  in cli-main [boot.janet] on line 3618, column 39

Following a lead from bakpakin, I tried disabling a part of this line: https://github.com/janet-lang/jpm/blob/cf6086ae3b282046141fd7b6fbd1d08edcb85fdd/jpm/shutil.janet#L101

A subsequent execution of jpm deps completed successfully.

Version info of various bits:

• Git for Windows 2.33.0.2 64-bit
• Janet 1.17.1 (from windows x64 msi installer)
• jpm (cf6086ae3b282046141fd7b6fbd1d08edcb85fdd)

[bakpakin]

Should be fixed in 588485be0c834d39ade9a967f254f9ca3a373082

[sogaiu]

I tried with 588485be0c834d39ade9a967f254f9ca3a373082 and 996a59e28ff758de5b73822dafa8e343e9879958 but still get similar output.

Before trying I tried to remove jpm via jpm uninstall from a local source repository for jpm and I checked the typical installation location here (C:\Users\user\AppData\Local\Apps\Janet\Library/jpm and C:\Users\user\AppData\Local\Apps\Janet\bin) to veritfy there the jpm-related files were not there.

I also verified after uninstalling jpm that invoking jpm resulted in the mesage 'jpm' is not recognized as an internal or external command, operable program or bathc file.

I also checked after pulling the latest jpm and installing it that shutil.janet had the changes in 588485be0c834d39ade9a967f254f9ca3a373082

There may still be pilot error here but if there is I haven't figured out what it is.

[sogaiu]

@bakpakin When I evaluate (os/environ) via jpm debug-repl, the resulting table has a key Path, but not a key for PATH. Perhaps that is to be expected.

When the merge-into in shell completes, the resulting table has both an entry with key Path and an entry with key PATH -- and the corresponding values differ.

I don't know what's going on with the subsequent os/execute, but I get the error output mentioned originally.

As a test, I changed the PATH key to Path for the struct being merged in. This results in error-less output.

I have tried examining the keys in the table that os/environ returns in different janet repl invocations and it appears that the key that ends up there can vary (i.e. sometimes it is Path and sometimes it is PATH).

[sogaiu]

Curiously, merging in a struct with both Path and PATH didn't seem to work here (same error message).

Checking the return value of os/environ to see if Path or PATH is already present and using whichever key was there (or if neither was there, just using PATH), seems to work:

  (def os-env (os/environ))
  (def path-key
    (if (get os-env "Path")
      "Path"
      # "PATH" keys exists or neither "Path" nor "PATH" keys exist
      "PATH"))
  (def env (merge-into os-env {"JANET_PATH" (dyn:modpath)
                               path-key (patch-path (os/getenv "PATH"))}))

(I replaced: https://github.com/janet-lang/jpm/blob/996a59e28ff758de5b73822dafa8e343e9879958/jpm/shutil.janet#L100-L101 with the eariler code in my successful test.)

[sogaiu]

Thanks for the change in 1cd77f978afab5c722e61c6f3cf8adb1dfbafa90 -- that worked here.