aegir 9.2.0 is failing in a weird way

[daviddias]

When running ipfs-bitswap tests with aegir 9.2.0:

• npm run test ✔︎
• npm run release 𝙓

?

It is a very strange behaviour. aegir 9.1.2 is 👌🏽

[dignifiedquire]

Your description is a little vague, could you list the exact errors, problems you are having?

[dignifiedquire]

Hours of debugging later

1. aegir@9.2.0 breaks some tests, including bitswap tests on js-ipfs
2. git bisect tells me the source of this is in 1d82ce7
3. Commenting out some code leads me to this line: https://github.com/dignifiedquire/aegir/blob/master/tasks/docs/publish.js#L4 which is just a require('gh-pages') call
4. Down the rabbit hole we go, commenting out requires in gh-pages/lib/index.js, leads me to this line: https://github.com/tschaub/gh-pages/blob/master/lib/index.js#L8
5. Looking at gh-pages/lib/git.js leads me to this line https://github.com/tschaub/gh-pages/blob/master/lib/git.js#L6 which requires q-io/fs
6. The source code of q-io/fs looks harmless, but I track the error down to this require: https://github.com/kriskowal/q-io/blob/v1/fs-mock.js#L7
7. I land on the repo for collections: https://github.com/montagejs/collections/blob/master/set.js
8. Turns out collections is modifying globals, including messing with the Array prototype. And that immediately on require 😢
9. Issues is known but not fixed, or doesn't seem to be planned to be fixed: Shims breaks crypto montagejs/collections#162

[dignifiedquire]

Opened kriskowal/q-io#160 so that this module hopefully won't be used in the future anymore. In the meantime I will look into either PRing a change to gh-pages or alternative modules.