You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Looking at the node-sass failures, it appears to be because the pre-built binary isn't compatible anymore. I'm guessing this is because the ABI changed but the module version didn't?
As for why it's erring the way it is, there looks to be a problem with our install script. After pulling down a prebuilt binary we try to execute the binary with a sample input. If an exception is raised we know the binary is no good. It looks like the node: symbol lookup error: error may not be catchable?
Because of the wired state the abi gets in (eg not bumped when it has
clearly changed) we're going to see this with any prebuilt or binary
bundled package because the abi version has broken it's stable api
guarantee. I think the version should be bumped as soon as it changes
because that will allow a graceful recompile is the "right thing". But I
may have to make that argument in a different venue.
Forcing a compile will catch incompatibilities with newer versions of node
and since that's a goal of the citgm I think it should be default.
Looking at the node-sass failures, it appears to be because the pre-built
binary isn't compatible anymore. I'm guessing this is because the ABI
changed but the module version didn't?
As for why it's erring the way it is, there looks to be a problem with our
install script. After pulling down a prebuilt binary we try to execute
the binary with a sample input. If an exception is raised we know the
binary is no good. It looks like the node: symbol lookup error: error may
not be catchable?
so it seems like the rebuild job is never getting to the "rebuild" lifecycle stage on node-sass as the install lifecycle hook is running a custom script, and it never has the chance to rebuild. The new release with bumped module version number should fix this.
As for why rebuild is not the standard approach... we want to actually catch things breaking as people would normally use npm, we can then use rebuild as a follow up for deeper diving. totally open to reviewing that process.
Sorry y'all it took me a good to while to figure out what's going here with node-sass. The issue stems from node-sass being installed from github rather than the npm. The github repo git submodules in LibSass. Since the submodule isn't being fetched CI, make is failing to find the LibSass sources files.
I spent some time today getting node-sass into a state where the npm package is compatible with node-citgm. I've have just published node-sass@v3.10.1. I have also opened #188 to remove the master flag from node-sass. 🚀
Activity
targos commentedon Sep 27, 2016
Related: #179
I'm pretty sure this is because we have to update
NODE_MODULE_VERSIONor force anode-gyp rebuildxzyfer commentedon Sep 27, 2016
Is there a Travis build I can look at?
MylesBorins commentedon Sep 27, 2016
Here is a citgm run with a forced rebuild
https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/396//cc @jasnell can you do an updated release of v7 with the NODE_MODULE_VERSION bumped?
edit: lolol no node-gyp... one more time
https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/397/
xzyfer commentedon Sep 27, 2016
https://ci.nodejs.org is currently 502'ing. I'll keep an eye on it.
xzyfer commentedon Sep 27, 2016
@thealphanerd sorry, I'm failing to find the node-sass build log.
MylesBorins commentedon Sep 27, 2016
bah... the rebuild job is not working in the regular CI job... made a custom CI job for rebuild, this should hopefully work
https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker-rebuild/2/
MylesBorins commentedon Sep 27, 2016
@xzyfer
xzyfer commentedon Sep 27, 2016
@thealphanerd is there a link to the build?
jasnell commentedon Sep 27, 2016
Yep. The hope is to get that updated build today.
On Monday, September 26, 2016, Myles Borins notifications@github.com
wrote:
reconbot commentedon Sep 27, 2016
Why doesn't the regular job rebuild? Seems like the right thing to do.
On Tue, Sep 27, 2016, 8:47 AM James M Snell notifications@github.com
wrote:
xzyfer commentedon Sep 27, 2016
Looking at the node-sass failures, it appears to be because the pre-built binary isn't compatible anymore. I'm guessing this is because the ABI changed but the module version didn't?
As for why it's erring the way it is, there looks to be a problem with our install script. After pulling down a prebuilt binary we
tryto execute the binary with a sample input. If an exception is raised we know the binary is no good. It looks like thenode: symbol lookup error:error may not be catchable?https://github.com/sass/node-sass/blob/master/scripts/build.js#L208-L218
reconbot commentedon Sep 27, 2016
Because of the wired state the abi gets in (eg not bumped when it has
clearly changed) we're going to see this with any prebuilt or binary
bundled package because the abi version has broken it's stable api
guarantee. I think the version should be bumped as soon as it changes
because that will allow a graceful recompile is the "right thing". But I
may have to make that argument in a different venue.
Forcing a compile will catch incompatibilities with newer versions of node
and since that's a goal of the citgm I think it should be default.
On Tue, Sep 27, 2016, 9:52 AM Michael Mifsud notifications@github.com
wrote:
MylesBorins commentedon Sep 27, 2016
so it seems like the rebuild job is never getting to the "rebuild" lifecycle stage on
node-sassas theinstalllifecycle hook is running a custom script, and it never has the chance to rebuild. The new release with bumped module version number should fix this.As for why rebuild is not the standard approach... we want to actually catch things breaking as people would normally use npm, we can then use rebuild as a follow up for deeper diving. totally open to reviewing that process.
reconbot commentedon Sep 27, 2016
I agree with you, I think getting the ABI version bumped as soon as it's broken is a good thing. =)
I also think your current process is 👍
MylesBorins commentedon Sep 27, 2016
I've submitted a PR that bumps the NODE_MODULE_VERSION on master.
nodejs/node#8808
This should
xzyfer commentedon Sep 28, 2016
node-sass tests are passing with the latest OSX nightly (node-v7.0.0-nightly201609276b443d1731-darwin-x64).
MylesBorins commentedon Sep 29, 2016
We've landed the bump to NODE_MODULE_VERSION
New ci to see what's up
https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/400/
targos commentedon Sep 29, 2016
node-sass failed:
Details
Use latest dist-tag for node-sass
xzyfer commentedon Sep 30, 2016
Sorry y'all it took me a good to while to figure out what's going here with node-sass. The issue stems from node-sass being installed from github rather than the npm. The github repo git submodules in LibSass. Since the submodule isn't being fetched CI, make is failing to find the LibSass sources files.
I spent some time today getting node-sass into a state where the npm package is compatible with node-citgm. I've have just published node-sass@v3.10.1. I have also opened #188 to remove the
masterflag from node-sass. 🚀Use latest dist-tag for node-sass
Use latest dist-tag for node-sass