Update page-lifecycle and y-indexeddb with tarball links #3262

codebuild-ro · 2025-10-02T07:16:07Z

Fixes #3021

I assumed the reason is because of upgraded yarn version.
So I replcaed git dependency url to tarball link.

page-lifecycle - from git+https://github.com/magic-akari/page-lifecycle#feat/add-types to https://codeload.github.com/magic-akari/page-lifecycle/tar.gz/50b50421bdeab3d211a57e81a277f699638373b0
y-indexeddb - from https://github.com/raineorshine/y-indexeddb#y-indexeddb-multiplex to https://codeload.github.com/raineorshine/y-indexeddb/tar.gz/60b960009085b1a988b5064ee35703229231531f

This will work because

Yarn downloads a pre-packed .tar.gz directly from GitHub’s codeload service instead of trying to git clone + pack.
No race conditions with tmp directories.
Dependabot’s environment handles tarball URLs consistently.

And regarding other updates, I reverted all of them.

Also I added codeload.github.com to allowed host list.

raineorshine · 2025-10-02T14:54:31Z

Could you explain what switching to tarball fixes and why? Thanks!

Also please link to the related GitHub issue as usual.

codebuild-ro · 2025-10-06T12:32:47Z

Could you explain what switching to tarball fixes and why? Thanks!

Here are the parts of error log of dependabot.

https://github.com/cybersemics/em/actions/runs/18173482130/job/51733584164#step:3:4666

https://github.com/cybersemics/em/actions/runs/18173482130/job/51733584164#step:3:5918

The failing part starts in Fetch step

Fetch Step
- Yarn decides to run npm pack to produce a .tgz because the repo has a package-lock.json.
- Same logic applies to y-indexeddb.
Tarbal Created
- npm pack actually created the archive page-lifecycle-0.1.2.tgz.
- Same for y-indexeddb-9.0.11-multiplex.0.tgz.
Renaming Step Fails
- Yarn expects to rename the tarball into a generic package.tgz.
- But by the time it tries, the file isn’t there anymore (or temp sandbox cleaned it up).
- Same error for y-indexeddb.
As the result, dependabot fails.

Tarball URLs fix this

If we replace the git+https://…#commit=… dependency with a direct .tar.gz URL (e.g. https://codeload.github.com/user/repo/tar.gz/), Yarn doesn’t need to:

Clone the repo,
Bootstrap with npm,
Pack into .tgz,
Rename into cache.

It simply downloads the tarball directly, which avoids the fragile temp directory + rename step that’s failing here.

raineorshine

Great, thank you!

raineorshine

I ran the CI a few more times, and it looks like Puppeteer is failing about 50% of the time:

Puppeteer is passing consistently in main.

This reverts commit 2e5cec6.

This reverts commit b0b2ff5.

This reverts commit b89243a.

raineorshine · 2025-10-15T17:59:33Z

Please provide greater transparency with your debugging process. What are your hypotheses? What clues do you have?

5. What is different about those test cases?

I am investigating what are the core differences now. But the same thing is that all 3 failing test cases are the first test case in each file.

If you change the order of the test cases (i.e. swap the first and the second test case), which one fails?

codebuild-ro · 2025-10-15T18:52:52Z

Please provide greater transparency with your debugging process. What are your hypotheses? What clues do you have?
If you change the order of the test cases (i.e. swap the first and the second test case), which one fails?

The first test always fails and if we swap first and second test cases, then second will be fail.

I finally find the reason of that failure.

In this `Serve` step in puppeteer flow, the server will be ready after a little tiny delay sometimes. But the puppeteer test will be started to run, even though the server is not ready. That makes some first test cases failed.

So I am trying to add waiting functionality until the server gets ready.

raineorshine · 2025-10-15T18:59:39Z

Please provide greater transparency with your debugging process. What are your hypotheses? What clues do you have?
If you change the order of the test cases (i.e. swap the first and the second test case), which one fails?

The first test always fails and if we swap first and second test cases, then second will be fail.

Thank you, that confirms that it is not the test order.

I finally find the reason of that failure.

In this Serve step in puppeteer flow, the server will be ready after a little tiny delay sometimes. But the puppeteer test will be started to run, even though the server is not ready. That makes some first test cases failed.
So I am trying to add waiting functionality until the server gets ready.

Great! I'll run it a few times to confirm.

raineorshine · 2025-10-15T19:55:38Z

package.json

-    "page-lifecycle": "git+https://github.com/magic-akari/page-lifecycle#feat/add-types",
+    "page-lifecycle": "^0.1.2",


We actually need the #feat/add-types branch from GoogleChromeLabs/page-lifecycle#12. v0.1.2 is missing type definitions.

Would you check this part please?
I just replace with the version from the code.
https://github.com/magic-akari/page-lifecycle/blob/50b50421bdeab3d211a57e81a277f699638373b0/package.json#L3
And real reference is declared in resolution object.
This means the modules will be fetch from tarball link.

"resolutions": { "@pandacss/node@npm:0.47.0": "patch:@pandacss/node@npm%3A0.47.0#~/.yarn/patches/@pandacss-node-npm-0.47.0.patch", "page-lifecycle": "https://codeload.github.com/codebuild-ro/page-lifecycle-fork/tar.gz/d407c14451b366fa156f53053bb42ab05d60719a", "y-indexeddb": "https://codeload.github.com/raineorshine/y-indexeddb/tar.gz/60b960009085b1a988b5064ee35703229231531f" },

Ahh, I forgot about resolutions! Do you know of any way to make this clearer for other developers? They might make the same mistake as me and not realize that it doesn't use the version listed in the package.json.

If resolutions overrides the package source, can we put any value there?

Do you know of any way to make this clearer for other developers?

What do you think about add special explanation in readme?

If resolutions overrides the package source, can we put any value there?

Regardless of what we put in dependencies, the resolutions value will be used.

The README is too far away for it to be useful for this kind of thing. It needs to be co-located.

If the dependency value is ignored, I would recommend putting a dummy value that points to resolutions. I'm not sure if this is very conventional, but I can't think of a better way. It would be misleading otherwise.

What about removing values from resolution and use direct value into dependency values?
https://github.com/cybersemics/em/pull/3262/files#diff-7ae45ad102eab3b6d7e7896acd08c427a9b25b346470d7bc6507b6481575d519R114
https://github.com/cybersemics/em/pull/3262/files#diff-7ae45ad102eab3b6d7e7896acd08c427a9b25b346470d7bc6507b6481575d519R148

That could work, though that just obscures it in the other direction, making it hard to know where the actual source code for the dependencies lives.

raineorshine · 2025-10-15T19:56:32Z

package.json

-    "y-indexeddb": "https://github.com/raineorshine/y-indexeddb#y-indexeddb-multiplex",
+    "y-indexeddb": "^9.0.11-multiplex.0",


Can you clarify this change? I'm not sure what ^9.0.11-multiplex.0 is. I don't remember https://github.com/raineorshine/y-indexeddb/tree/y-indexeddb-multiplex getting merged into the official repo.

This is also same reason with page-lifecycle for version and tarball link.
https://github.com/raineorshine/y-indexeddb/blob/60b960009085b1a988b5064ee35703229231531f/package.json#L3

raineorshine

Thanks!

This looks good if it solves the original issue. We just need a way to document and link developers to the source of the dependencies, as the tarball url currently obscures where it comes from.

See: #3262 (comment)

.github/workflows/puppeteer.yml

…th 10 seconds always

codebuild-ro · 2025-10-17T04:50:17Z

.github/workflows/puppeteer.yml

+
+            # Test the full request path and check for 502 errors
+            HTTP_STATUS=$(curl -s -o /dev/null -w "%{http_code}" -k https://localhost:2552 2>/dev/null || echo "000")
+


When start Test step, the log said that the server responds but with 502 (Bad gateway) status.
So I added script to wait until the server returns with 200 status code.

raineorshine · 2025-10-17T16:17:40Z

@karunkop Could you confirm that this change doesn't introduce any flakiness by running the CI 10 times? Thanks!

karunkop · 2025-10-22T02:09:26Z

@karunkop Could you confirm that this change doesn't introduce any flakiness by running the CI 10 times? Thanks!

This has been tested in CI sequentially for 30 times, and I can confirm that there is no flakiness seen in this PR. Here is the result of the checks.

Add git based dependencies into .ncurc configuration

2150149

codebuild-ro marked this pull request as draft October 2, 2025 07:19

codebuild-ro marked this pull request as ready for review October 2, 2025 07:42

codebuild-ro self-assigned this Oct 2, 2025

codebuild-ro marked this pull request as draft October 2, 2025 07:54

codebuild-ro force-pushed the fix/re-enable-dependabot branch from 65c54e7 to 45f4f4d Compare October 2, 2025 08:07

Update page-lifecycle and y-indexeddb with tarball

431c8da

codebuild-ro force-pushed the fix/re-enable-dependabot branch from 45f4f4d to 431c8da Compare October 2, 2025 08:34

codebuild-ro closed this Oct 2, 2025

codebuild-ro deleted the fix/re-enable-dependabot branch October 2, 2025 08:40

codebuild-ro restored the fix/re-enable-dependabot branch October 2, 2025 08:40

codebuild-ro reopened this Oct 2, 2025

Merge branch 'fork_main' into fix/re-enable-dependabot

97874bf

codebuild-ro marked this pull request as ready for review October 2, 2025 10:54

Merge branch 'fork_main' into fix/re-enable-dependabot

40cbe6a

raineorshine approved these changes Oct 6, 2025

View reviewed changes

raineorshine requested changes Oct 6, 2025

View reviewed changes

codebuild-ro force-pushed the fix/re-enable-dependabot branch 3 times, most recently from 9805466 to 40cbe6a Compare October 7, 2025 09:15

codebuild-ro added 2 commits October 7, 2025 11:17

Merge branch 'fork_main' into fix/re-enable-dependabot

5eeece7

Temp Testing for failing puppeteer: rollback page-lifecycle

2e5cec6

codebuild-ro force-pushed the fix/re-enable-dependabot branch from d1f6770 to 2e5cec6 Compare October 7, 2025 09:32

codebuild-ro added 5 commits October 7, 2025 15:39

Revert "Temp Testing for failing puppeteer: rollback page-lifecycle"

5b14496

This reverts commit 2e5cec6.

Skip failing test case for testing

b0b2ff5

Revert "Skip failing test case for testing"

ecec300

This reverts commit b0b2ff5.

Increase timeout values

b89243a

Revert "Increase timeout values"

8e377d5

This reverts commit b89243a.

Temp Testing for failing puppeteer: rollback page-lifecycle

5c52db7

codebuild-ro force-pushed the fix/re-enable-dependabot branch from 3d58765 to 5c52db7 Compare October 15, 2025 08:35

codebuild-ro added 2 commits October 15, 2025 11:14

Fix page-lifecycle with tarball link

10e8808

update puppeteer.yml for let server runs correctly on CI

6d1674d

codebuild-ro force-pushed the fix/re-enable-dependabot branch from 6f7cca8 to 998a84a Compare October 15, 2025 18:30

codebuild-ro closed this Oct 15, 2025

codebuild-ro reopened this Oct 15, 2025

codebuild-ro force-pushed the fix/re-enable-dependabot branch from 998a84a to 3e1e52a Compare October 15, 2025 19:03

Add waiting func for server runs

bd87d4c

codebuild-ro force-pushed the fix/re-enable-dependabot branch from 3e1e52a to bd87d4c Compare October 15, 2025 19:11

codebuild-ro requested review from raineorshine October 15, 2025 20:01

raineorshine reviewed Oct 15, 2025

View reviewed changes

codebuild-ro added 4 commits October 16, 2025 07:45

Change to direct tarball link instead of using resolutions object

23fcf9a

Update yarn.lock

35d96a4

Update page lifecycle tarball with original repository

ce4cb89

Update yarn.lock

a125460

raineorshine requested changes Oct 16, 2025

View reviewed changes

.github/workflows/puppeteer.yml Outdated Show resolved Hide resolved

update puppeteer.yml to wait until server runs, instead of waiting wi…

99e792f

…th 10 seconds always

codebuild-ro commented Oct 17, 2025

View reviewed changes

update package.json and README.md for dependency link urls

da95bbe

raineorshine requested a review from karunkop October 17, 2025 16:17

raineorshine approved these changes Oct 17, 2025

View reviewed changes

karunkop approved these changes Oct 22, 2025

View reviewed changes

raineorshine merged commit d13dce4 into cybersemics:main Oct 22, 2025
20 of 22 checks passed

		"page-lifecycle": "git+https://github.com/magic-akari/page-lifecycle#feat/add-types",
		"page-lifecycle": "^0.1.2",

		"y-indexeddb": "https://github.com/raineorshine/y-indexeddb#y-indexeddb-multiplex",
		"y-indexeddb": "^9.0.11-multiplex.0",


		# Test the full request path and check for 502 errors
		HTTP_STATUS=$(curl -s -o /dev/null -w "%{http_code}" -k https://localhost:2552 2>/dev/null \|\| echo "000")

Update page-lifecycle and y-indexeddb with tarball links #3262

Update page-lifecycle and y-indexeddb with tarball links #3262

Uh oh!

Conversation

codebuild-ro commented Oct 2, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

raineorshine commented Oct 2, 2025

Uh oh!

codebuild-ro commented Oct 6, 2025

Tarball URLs fix this

Uh oh!

raineorshine left a comment

Choose a reason for hiding this comment

Uh oh!

raineorshine left a comment

Choose a reason for hiding this comment

Uh oh!

raineorshine commented Oct 15, 2025

Uh oh!

codebuild-ro commented Oct 15, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

raineorshine commented Oct 15, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

codebuild-ro Oct 16, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

raineorshine left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

raineorshine commented Oct 17, 2025

Uh oh!

karunkop commented Oct 22, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

codebuild-ro commented Oct 2, 2025 •

edited

Loading

codebuild-ro commented Oct 15, 2025 •

edited

Loading

codebuild-ro Oct 16, 2025 •

edited

Loading