std: fs/expand_glob.ts: expandGlob broken from 2c7274741ca669c3ee125623657e4f37e39408a0

[cscheid]

The commit 2c7274741ca669c3ee125623657e4f37e39408a0 seemingly broke ** expansion in expandGlob, in 0.161.0.

Repro:

$ mkdir repro
$ cd repro
$ mkdir -p foo/bar
$ touch 1
$ touch foo/1
$ touch foo/bar/1
$ deno
Deno 1.28.1
exit using ctrl+d, ctrl+c, or close()
> import { expandGlobSync } from "https://deno.land/std@0.160.0/fs/mod.ts";

undefined
> console.log([...await expandGlobSync("**/1")].length);
3
undefined
>
$ deno
Deno 1.28.1
exit using ctrl+d, ctrl+c, or close()
> import { expandGlobSync } from "https://deno.land/std@0.161.0/fs/mod.ts";

undefined
> console.log([...await expandGlobSync("**/1")].length);
2
undefined
>

[cscheid]

@bartlomieju I apologize for posting to the wrong repo; I thought for a second you were tracking everything over there. It's possible I missed a duplicate here as well, then.

[kamilogorek]

The globstar is not enabled by default for expandGlob and expandGlobSync, which means that to retain the same behavior between 0.160 and 0.161 you need to call it as expandGlobSync("**/1", { globstar: true })

[cscheid]

Thanks for the followup. Still, the documentation itself states that the default behavior should be that of globstar: true:

boolean = true, and "if false [emphasis mine], ..." both imply that that by default, ** should behave like *.

Let me also mention that the release notes do not indicate behavior change: https://github.com/denoland/deno_std/releases/tag/0.161.0, and neither do the PR or the issue.

[kamilogorek]

You are looking at path module, not fs, but because one inherits from other, I agree that it should be in sync. Filled a PR for that.