[PDE-4679] fix(cli): better handling of broken symlinks while copying files to temp directory during build process

[rnegron]

TO-DO: This is still missing some tests.

Intended to fix #706. Modifies the copyDir method by using lstat to avoid crashing when reading a faulty symbolic link.

About lstat from the manual pages:

Example of new warning message:

[eliangcs]

zapier build produces two files:

• build/build.zip
• build/source.zip.

source.zip already respects .gitignore. However, I don't think build.zip should exclude the files in .gitignore. When running zapier build --skip-npm-install, we intentionally don't run npm install internally. Instead, it copies the files from the working directory's node_modules. So even though node_modules is in .gitignore, we should include it in build.zip.

I believe the problem with #706 is not about .gitignore but is about invalid symlinks. zapier build should show a better error message when it encounters a bad symlink. Let me know if I missed anything though!

Also, be sure to test different options:

• zapier build --skip-npm-install (currently crashing with this PR)
• zapier build --disable-dependency-detection

[rnegron]

@eliangcs

So even though node_modules is in .gitignore, we should include it in build.zip.

OK, this makes sense to me! I was wondering if just using gitignore was the correct approach or not. I will re-tool this PR to add some logic for detecting an invalid symlink and erroring appropriately.

[rnegron]

Hmm, using lstat causes the describe('build in workspaces') tests to fail. Checking...

Update: the issue is not lstat but rather the fact that yarn workspaces creates symlinks to directories such as: /private/var/folders/28/81v0tq5s2cv9zcnvhgtrr9jw0000gn/T/zapier-90e1abb9/node_modules/app-1 that point to ../packages/app-1, and symlinks-to-directories are not supported yet. Will fix tomorrow.

[rnegron]

I was thinking that, in order to address the original issue, simply alerting the user that the problem was a broken symlink might be enough and we don't need to get too creative by differentiating between file or directories...

The question is then: should we skip the faulty symlink or error out of the build process?

[eliangcs]

The question is then: should we skip the faulty symlink or error out of the build process?

I think skipping with a very visible warning (like you're doing now) would be good.

This looks good to me. Your comment reminds me that we need to skip the symlinks that point to a workspace. I'll go ahead and merge this, so I can fix that in my PR. Thanks!