Giters

nodejs / citgm

Canary in the Gold Mine

Home Page:https://www.npmjs.com/package/citgm

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Latest CITGM results

RafaelGSS opened this issue · comments

commented

Let's use this issue to list failing modules and take action such as removing offending modules, skip specific arches and so on

@nodejs/citgm

commented

tape failed on rhel8-ppc64le due to a timeout; i suspect with a longer timeout it'd pass.

commented

I just had a nice chat with @RafaelGSS. Will look into it :)

commented

Regarding this test fail.
On my machine works fine with these specifications:

  • Node v20.8 (as in the CI run)
  • underscore-v1.13.6 (as in the CI run)
  • macOs version 13.6

Command run: citgm underscore@1.13.6
The tests are not flacky on this package.

commented

All the issues regarding macOS are not reproducible for me locally. I'm using the correct Node version and the correct package version, the tests are not failing on my machine. @nodejs/build

commented

Let's wait v18.x and main to see if the same happens for these versions. But, I guess that's something related to our machines.

commented

The failures on three platforms for nan already has an issue (#996). It would be great if someone could figure that one out. (I haven't looked and don't have the right expertise probably anyway.)

EDIT: Oh, I see @RafaelGSS already linked to it in the description. I'd love to see someone get to the bottom of that one.

commented

The macOS failures have a PR that might be worth landing just to bypass things for straightforward JavaScript modules.

#971

commented

tape failed on rhel8-ppc64le due to a timeout; i suspect with a longer timeout it'd pass.

@ljharb Is it possible to pass a longer timeout value via the command line? Something along the lines of npm run tests-only -- --timeout=99999? If so, it would just be a matter of updating line 483 of lookup.json.

citgm/lib/lookup.json

Line 483 in cd1b288

"scripts": ["tests-only"],

commented

Is really interesting to notice that in the last run on main there are no issues over macOS even without the skip applied, but on v18.18 there are 6 errors.

I managed to reproduce the issue over pino@8.15.3 and also the multer issue. The multer one is flaky on my machine, here is the PR to fix it

commented

@ljharb probably also semver failed due to a timeout. Shall we try to increase it or set it as flaky due to tests timeout?

commented

Please keep the max timeout to a reasonable amount. We don't want to end up with a citgm run that takes forever to timeout when something got stuck in the test.

commented

I managed to reproduce the issue over pino@8.15.3 and also the multer issue. The multer one is flaky on my machine, here is the PR to fix it

While that's a good short-term solution to the problem, an actual fix should hopefully land in either multer or node so that it stop failing on macos. In these cases please open an issue pinging the maintainer - you can find it by browsing lookup.json e.g:

citgm/lib/lookup.json

Lines 337 to 341 in cd1b288

"multer": {
"prefix": "v",
"skip": "win32",
"maintainers": "linusu"
},

Here's an example of an issue pinging the maintainer of a package for help: #988. Keep in mind you might be able to post more info, since you're able to reproduce in your machine you might be able to add some helpful logs or maybe test which combinations of package version + node version it fails vs succeeds, etc.

Thanks for the help @alfonsograziano it's really appreciated! 🙏

commented

Here are a few issues/PRs open now to fix these issues:

commented

After the latest run on v18.18, we have 14 issues.

  • torrent-stream: we have already #1001 to fix this issue. The package will be removed as required from the maintainer in #1000. @RafaelGSS can you please merge the PR? :)
  • multer: the package is still failing on 3 platforms including macOS, even after #998 has been merged. I have no idea why is still failing. If continue to fail, shall we make it flaky on all the platforms that are failing now?
  • nan: still failing, I proposed to skip on the failing platforms on #996. Addressed in #1005
  • leveldown: install failed (macOS) #1006
  • microtime: install failed (macOS) #1006
  • bufferutil: install failed (macOS) #1006
  • underscore: 2 tests failed on macOS, cannot reproduce on my mac. #1008
  • pino: Discussion on #1002
  • semver: failed for timeout, issued #1004 to fix it
commented

After the latest run on v21.0.0, we have 14 issues.

  • nan: as above, addressed in #1005
  • torrent-stream: as above, addressed in #1001
  • express-session: 1 test failure
  • jest: ENOSPC: no space left on device. Do we know how to fix it? @RafaelGSS
  • import-in-the-middle: Discussion happening in #995
commented

For the install failing, I think we should just drop these packages if we confirm these failures aren't machine-related.

jest: ENOSPC: no space left on device. Do we know how to fix it? @RafaelGSS

Usually, we open an issue on nodejs/build.

commented

For the install failing:

They are always failing on v18.18 but I cannot reproduce the failure on my local machine.
#1006 will skip them for now on OSX and at the same time will create an issue to find a proper solution :)

Usually, we open an issue on nodejs/build.

Done here:
nodejs/build#3512
Outcome: The space has been increased by deleting some files, we now have more than 5gb free. Hopefully shouldn't happen in the next run

commented

@RafaelGSS I'm not sure if this run is correct, there are still errors from torrent-stream that has been removed entirely from the lookup

commented

@RafaelGSS I'm not sure if this run is correct, there are still errors from torrent-stream that has been removed entirely from the lookup

The CI job runs by default on published versions of CITGM -- there hasn't been a release with the changes to the lookup.

commented

@RafaelGSS I'm not sure if this run is correct, there are still errors from torrent-stream that has been removed entirely from the lookup

The CI job runs by default on published versions of CITGM -- there hasn't been a release with the changes to the lookup.

Should I use nodejs/citgm#main on
image
?

UPDATE: Running https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/3318

commented

The current status after this run is: 15 failures on these packages:

  • import-in-the-middle: failed on all the platforms. Asked for maintainers support in #995
  • ava: failed on ubuntu, debian. Asked for maintainers support in #1012. I'm a little concerned by the fact that this module is already skipped on a lot of platforms, maybe the issue that is triggered on the two platforms can affect other platforms as well.
  • jose: failed on ubuntu, rhel8-s390x, rhel8-ppc64le, debian10-x64, rhel8-x64. Asked for maintainers support in #1011
  • semver: failed on rhel8-ppc64le - increase timeout proposed in #1013. Not sure if this is the best solution to the problem
  • jest: failed on debian - test timed out after 33 minutes. I see that in the lookup we have a 30 mins timeout. Is there any reason for such a long timeout on this module?
commented

I've traced the issue down to a recent zlib upgrade that landed on main which changed the outputs. Opened nodejs/node#50138 and requested a revert and further investigation.

commented

It seems we got a green CI for LTS release lines. I'm opening #1033 for v21 failures.