v22.11.1 intrododuces publishing bug

[mixmix]

Problem

Node: 14.16.1
OS: Linux, MacOS

have noticed the following error with publishing builds with v22.11.1 :

> electron-builder --config build/electron-builder.config.js --publish always

  • cannot check updates  error=TypeError: update_notifier_1.default is not a function
  • electron-builder  version=22.10.5 os=5.4.0-72-generic
  • loaded configuration  file=/home/somename/projects/AHAU/whakapapa-ora/desktop/build/electron-builder.config.js
  • writing effective config  file=dist/installers/builder-effective-config.yaml
  • rebuilding native dependencies  dependencies=fd-lock@1.2.0, fsctl@1.0.0, leveldown@5.6.0, sodium-native@3.2.1, utp-native@2.4.0 platform=linux arch=x64
  • packaging       platform=linux arch=x64 electron=12.0.6 appOutDir=dist/installers/linux-unpacked
  • building        target=AppImage arch=x64 file=dist/installers/ahau-desktop-Linux-1.3.1-x86_64.AppImage
  • publishing      publisher=Github (owner: protozoa-nz, project: whakapapa-ora, version: 1.3.1)
  • uploading       file=ahau-desktop-Linux-1.3.1-x86_64.AppImage provider=GitHub
  ⨯ Cannot read property 'getType' of undefined  failedTask=build stackTrace=TypeError: Cannot read property 'getType' of undefined
    at GitHubPublisher.doUploadFile (/home/somename/projects/AHAU/whakapapa-ora/desktop/node_modules/electron-publish/src/gitHubPublisher.ts:190:36)
    at GitHubPublisher.doUpload (/home/somename/projects/AHAU/whakapapa-ora/desktop/node_modules/electron-publish/src/gitHubPublisher.ts:170:23)
    at processTicksAndRejections (internal/process/task_queues.js:93:5)
    at GitHubPublisher.upload (/home/somename/projects/AHAU/whakapapa-ora/desktop/node_modules/electron-publish/src/publisher.ts:87:5)
    at async Promise.all (index 1)
    at AsyncTaskManager.awaitTasks (/home/somename/projects/AHAU/whakapapa-ora/desktop/node_modules/builder-util/src/asyncTaskManager.ts:65:25)
    at PublishManager.awaitTasks (/home/somename/projects/AHAU/whakapapa-ora/desktop/node_modules/app-builder-lib/src/publish/PublishManager.ts:228:5)
    at Object.executeFinally (/home/somename/projects/AHAU/whakapapa-ora/desktop/node_modules/builder-util/src/promise.ts:23:3)

Investigation

Looked at /home/somename/projects/AHAU/whakapapa-ora/desktop/node_modules/electron-publish/src/gitHubPublisher.ts:190:36 and it's saying mime is undefined which is pretty strange

➜  desktop git:(master) npm ls mime --depth 5
ahau-desktop@1.3.1 /home/somename/projects/AHAU/whakapapa-ora/desktop
├─┬ electron-builder@22.11.1
│ └─┬ app-builder-lib@22.11.1
│   └─┬ electron-publish@22.11.1
│     └── mime@2.5.2 deduped

hmmm

Fix

regress to v22.10.5

[cawa-93]

I can confirm for Windows: https://github.com/cawa-93/vite-electron-builder/runs/2507942045

[modmoto]

@mixmix
What tool did you downgrade? We tried electron-publish but that did not work

edit:
found the issue, just adding "electron-builder": "22.10.5" to the "devDependencies" worked.

[mixmix]

Electron-builder

…

On Thu, 6 May 2021, 22:52 s.heiss, ***@***.***> wrote: @mixmix <https://github.com/mixmix> What tool did you downgrade? We tried electron-publish but that did not work — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#5861 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAUK3HUIZHHWHBWII55EU7TTMJYF3ANCNFSM44F4F4ZQ> .

[cawa-93]

Ping @develar @mmaietta

[mixmix]

BTW, I'm totally happy using the earlier version. I'm incredibly grateful for the work that's been put into this module. It's making building traditional knowledge management software for Māori communities in New Zealand. Opened this issue just for visibility

…

On Sat, 8 May 2021, 22:47 Alex Kozack, ***@***.***> wrote: Ping @develar <https://github.com/develar> @mmaietta <https://github.com/mmaietta> — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#5861 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAUK3HTW3X2UIGJKVH5E6KLTMUJDTANCNFSM44F4F4ZQ> .

[mmaietta]

There were a lot of changes between 22.10.5 to 22.11.1, which makes this so much more difficult to triage.
Few changes that I've collected so far that could potentially relate to this.
ts-babel -> tsc
yarn -> pnpm

    "target": "es2018", -> es2019
    "module": "esnext", -> CommonJS

I have no idea how to test to fix this though as I don't have any publisher setups to work with.

Could someone please put together a minimal-reproducing repo/gist?

[cawa-93]

There were a lot of changes between 22.10.5 to 22.11.1, which makes this so much more difficult to triage.
Few changes that I've collected so far that could potentially relate to this.
ts-babel -> tsc
yarn -> pnpm

    "target": "es2018", -> es2019
    "module": "esnext", -> CommonJS

I have no idea how to test to fix this though as I don't have any publisher setups to work with.

Could someone please put together a minimal-reproducing repo/gist?

You can use my template repo, if this help ...
https://github.com/cawa-93/vite-electron-builder/

[bcomnes]

Could someone please put together a minimal-reproducing repo/gist?

I wrote up the problem in the published package:

#5870 (comment)

Basically, the tests are treating module.exports in mime as a default import, but is treating module.exports as an object of named exports and looking for a default property on that object in the published asset. Since mime doesn't export an object with a default property, you will see these .default is not a function errors wherever a cjs module is imported as a default import.

The solution is:

• make the ts test and publish config consistent somehow (I have no idea)
• fix the import syntax to match the config

[mmaietta]

make the ts test and publish config consistent somehow (I have no idea)

I honestly have no idea either. I also don't have access to publishing the electron-builder package to npm, or know what scripts are running during/for the publish.
CC @develar for access 😉

One approach is to revert to ts-babel so we can isolate the pnpm migration from the compiler changes. Might be at least a safe route since so many changes occurred between 22.10.5 and 22.11.1, which seems to have led to a significant jump in tickets opened

[bcomnes]

I would bet this issue is on the tsc side and not on the pnpm side, since it seems like its a transpilation configuration issue, and not a module install/resolution issue.

[develar]

Fix: import * as mime from "mime" instead of import mime from "mime". Will be pushed soon.