This change uses a different JavaScript syntax to cause the "when a .js file
does not start with 'use babel'" spec to fail again after upgrading to Electron
3 / Node 10.3.0. The previous generator syntax seems to be supported in the
version of V8 used by Node 10 so we now need to use the 'export default' syntax
to trigger an error.
I've verified that changing the first line of this file to 'use babel' causes
the file to load successfully, so it seems this new syntax is effective to
verify the desired behavior.
This change uses a different JavaScript syntax to cause the "when a .js file
does not start with 'use babel'" spec to fail again after upgrading to Electron
3 / Node 10.3.0. The previous generator syntax seems to be supported in the
version of V8 used by Node 10 so we now need to use the 'export default' syntax
to trigger an error.
I've verified that changing the first line of this file to 'use babel' causes
the file to load successfully, so it seems this new syntax is effective to
verify the desired behavior.
* Remove the `legacyScopeName` field
* Remove the legacy scope name concept from the Config class
* Handle tree-sitter grammars and textmate grammars having the same
scope names
Prior to this change, these tests assumed that the local repository is a
clone of atom/atom, and the tests (unintentionally) failed if the local
repository was a fork of atom/atom:
https://circleci.com/gh/marcomorain/atom-1/29
Instead of depending on the local clone of the atom repository for these
tests, this commit updates the tests to use one of the fixture
repositories. 😅
In attempting to optimize the performance of `isFoldableAtBufferRow` in
3c87b74, we mistakenly introduced a bug that caused lines that contained
a comment but didn't start with one to not be foldable anymore.
With this commit we are restoring the previous behavior, thus only
disabling folding for lines that start with a comment (ignoring leading
whitespaces).
Previously, a guard based on the presence of the _id field (which is
inserted by npm during installation) prevented a regex replacement of
the git+ prefix on URLs. Now we always do this. Since the .git suffix
also causes problems and we’re removing that in packages, I now remove
that as well.