Closed phated closed 4 years ago
It could be a problem with the version of zip I am using. I have it pegged at 0.0.6, but looks like there is a 1.1.0 out. Will update to the latest. If you wanted to try something hacky in the meantime, you could go to the volo installation directory and manually force zip 1.1.10, maybe by modifying the package.json. No guarantees that hack will work though, not sure if it changed APIs or may require other updates.
I updated to the latest zip module in a fork and it passed all the tests. Then, I modified the branch I have been working in to use that fork and pushed to travis, which ran completely on both node 0.8.22 and 0.10.2 (ref build). The issue has typically been spotty, so I can't say that it is completely fixed. Though, It is a good sign and I will keep this updated on whether it fails again.
Fixed in 9304919d9b8bd2c3e53e1bbc97628508337156e3
Did some work on a plane and did not have access to github issues area, so could not mention the fix number in the commit.
This seems to still be happening with 0.3.0
If I were able to swap https://github.com/nearinfinity/node-unzip for the zip dependency, would that be okay? Bower seems to be using the unzip library.
Scratch that, the library that Bower uses has big problems on Node 0.10.
On another note, I can't get this error to occur on node 0.8.22 on my local machine, but it is still happening on travis - I guess I need to look into it some more.
I was having trouble getting all the unit tests to pass with node 0.10. The error seemed to be related to the http module, seemed to fail on a certain kind of request, but I could not reproduce when I tried each of the test volo add
s by themselves. Very frustrating.
I still feel like node and its ecosystem is adjusting to the changes in 0.10. It could also be a q interaction with 0.10 too.
In any case, when you run in travis, does it use node 0.8 or 0.10?
It would help in any of these cases if we could get something we could reproduce on our local dev machines too. That would decrease the frustration level if we could start to isolate what is causing the problem.
0.10 seems to always be working on travis. 0.8 is the one that keeps producing the error. I could see http being the culprit if it wasn't pulling down a zip correctly. Another note, this seems to be happening against github dependencies.
Huh, OK, well so much for the "0.10 seems uneven" theory, if 0.8 is failing on travis.
When you say "happens against github dependencies" do you mean fetching .zip files of github repos or something else? If .zip files of repos, then what kinds of .zip files do seem to consistently work?
Urls such as https://codeload.github.com/bestiejs/lodash/legacy.zip/master
We don't grab any other types of .zips.
This doesn't happen every time, but often enough.
When running
volo add
on travis-ci with node 0.8.22, I often get the messageERROR: Error: Unable to find end of central directory record
See https://travis-ci.org/iceddev/frozen/jobs/6893229 for an example.
I have tried modifying my repo in many ways to solve this problem but it seems to be a problem with the zip files. Not sure if something in volo is changing them or the Github API has changed or is being flakey.