-
Notifications
You must be signed in to change notification settings - Fork 34
Conversation
@@ -19,8 +19,6 @@ if [ "$TRAVIS_OS_NAME" = "osx" ]; then | |||
export ATOM_SCRIPT_PATH="./atom-${ATOM_CHANNEL}" | |||
ln -s "./atom/${ATOM_APP_NAME}/Contents/Resources/app/atom.sh" "${ATOM_SCRIPT_PATH}" | |||
fi | |||
export PATH="$PWD/atom/${ATOM_APP_NAME}/Contents/Resources/app/apm/bin:$PATH" | |||
export ATOM_PATH="./atom" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This wasn't being used in the script, so I removed it as well.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
ATOM_PATH
is used by Atom though, specifically atom.sh
.
Looks like I didn't explicitly mention this: This bug completely breaks linting with |
There is a node executable in the apm/bin directory, which overrides any version of Node.js installed on the system since the PATH is being updated to put this path at the front. As the Linux side of the script does not do this, remove the modification of the PATH to stop breaking Node.js.
@atom/feedback This seems to work OK for normal packages. Are there any edge cases that I'm missing? Since this seems to be breaking CI for a lot of packages, I'm tempted to 🚢 this. I'll wait a 12-ish hours unless I hear otherwise. After that I'm merging this 👶 |
I'm gonna merge this. If it breaks stuff, blame me 🔫 |
Turns out this won't fix any of the eslint problems since I'm assuming practically everyone has it listed as a dependency in their package's One workaround for now is to not rely on the automatic linting support provided here, but rather to install eslint manually using npm, and also performing the linting manually. Sucks, but that's the reality. The better workaround is to downgrade eslint of course, until atom/atom#11897 🚢s. |
👎 on merging this; I think it should be reverted. |
@joefitzgerald Why? We only set |
Sure, let's do it for all of them. We should be using bundled node to do everything if possible. If people have issues with that, then - well - we need to upgrade bundled node (work in progress via @BinaryMuse) because it's what is being used anyway. |
OK, as long as we have consistent behavior across all platforms, I agree that this can be reverted since it doesn't actually fix the intended problem. |
I would be quite surprised if many projects (besides How exactly do you propose we go about this then? Force everyone to wait on updating to |
@Arcanemagus it's not clear to me that merging the PR actually resolved the issue for you. Thus, isn't this line of reasoning (that eslint 3.x is blocked on this PR) potentially missing the root cause of the eslint issue? Regardless of |
It had solved it for a few things, but others are broken due to the issue outlined in #43 (comment). I think waiting on a stable release of Atom to include a fixed version of My best idea that I've had so far is:
Do you see anything wrong with that, other than the slightly increased CI run times? |
Reverts the changes from atom#43 per atom#43 (comment).
Reverts the changes from atom#43 per atom#43 (comment).
For anyone coming across this issue, I've implemented the above workaround over in https://github.com/Arcanemagus/ci/tree/atomlinter until an updated version of |
There is a node executable in the apm/bin directory, which overrides any version of Node.js installed on the system since the PATH is being updated to put this path at the front. As the Linux side of the script does not do this, remove the modification of the PATH to stop breaking Node.js.
You can see a simplified version of the script being run here. This shows that Node.js is v6.2.2 at the start of the script, but after the
PATH
modification it shows Node.js at v0.10.40.I'm not sure why this modification was there in the first place though, @joefitzgerald as you are the original author of that do you know if this is necessary?