fix(scripts): make it so prepack runs properly #419
Conversation
|
not sure why, but my tests aren't even passing locally (on main!) so it's hard to figure out what's actually going on here. |
|
is your local fork out of sync? there's a lot of weird diffs here I don't understand. |
|
oh, is it? Maybe that's it. One sec. |
|
ughhhh I totally messed that up. One sec. |
🤖 I have created a release *beep* *boop* --- ## [21.0.0](npm/pacote@v20.0.0...v21.0.0) (2024-11-25) ###⚠️ BREAKING CHANGES * `bun.lockb` files are now included in the strict ignore list during packing * this module is now compatible with the following node versions: ^20.17.0 || >=22.9.0 ### Bug Fixes * [`844dc08`](npm@844dc08) update node engines to ^20.17.0 || >=22.9.0 (npm#414) (@wraithgar) ### Dependencies * [`2cb6fa7`](npm@2cb6fa7) [npm#415](npm#415) `npm-packlist@10.0.0` (npm#415) * [`47b928c`](npm@47b928c) [npm#412](npm#412) replace node builtin rmSync with rimraf (npm#412) (@mbtools) ### Chores * [`b6f35a2`](npm@b6f35a2) [npm#402](npm#402) bump @npmcli/arborist from 7.5.4 to 8.0.0 (npm#402) (@dependabot[bot]) * [`1ef54ba`](npm@1ef54ba) [npm#408](npm#408) support tests on win32 (npm#408) (@mbtools) * [`555b000`](npm@555b000) [npm#401](npm#401) bump @npmcli/template-oss from 4.23.3 to 4.23.4 (npm#401) (@dependabot[bot], @npm-cli-bot) --- This PR was generated with [Release Please](https://github.com/googleapis/release-please). See [documentation](https://github.com/googleapis/release-please#release-please). Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
|
alright. I'm not sure what the two test failures are actually about: I've manually tested that this works, and I think I just don't remember how this part of tap works. I guess the test is hanging somehow? I also have NO idea what the other failure is about. Maybe they're related? Workspaces are just not a thing I ever touched. |
|
my access is gonna be cut off any minute now, so is it ok if I just leave this one in your hands? I'm pretty sure it's just about ready to get over the finish line. Although my brain's been really scrambled today. I might have just COMPLETELY missed the mark. In which case I apologize and please close and ignore. I was just trying to help the TS team with one last thing :) |
| @@ -54,7 +52,22 @@ class DirFetcher extends Fetcher { | |||
| npm_package_json: resolve(this.resolved, 'package.json'), | |||
| }, | |||
| }) | |||
| }) | |||
| } | |||
| if (process.env._PACOTE_FROM_GIT_ === 'yes' && mani.scripts?.prepack) { | |||
There was a problem hiding this comment.
I'm confused why this secret env is needed. These are generally a code smell.
There was a problem hiding this comment.
I chose that because it seemed to be the chosen strategy for configuring the underlying NPM subprocess with special behavior when the git fetcher is invoking the dir fetcher (see _PACOTE_NO_PREPARE_). There could be better ways, like checking the spec of the package being built to see if it's a git spec, though. Idk why I didn't think of that.
There was a problem hiding this comment.
I'll have to noodle on this. We do know at some point we are from git and could set real state here that's not process.env. Maybe.
Fixes: #257