add support for fetch timeout control for node_helpers, fix timeouts on armv6l

[sdetweil]

user reporting slow/no connection/timeout errors on armv6l for calendar, and newsfeed

we can increase the timeout by adding calls to the undici lib, but it requires node 20.18.1 or above.

this adds the support for timeout
(also environment variable to override if needed,, mmFetchTimeout (default 30 seconds)

and updates the base node version

[bugsounet]

Warn: package-lock

integrity/resolve missing

[khassel]

for changing the built-in node fetch timeout we have to add the whole undici http client as dependency?

I would prefer a simpler solution e.g. adding env var --network-family-autoselection-attempt-timeout (untestet)

[sdetweil]

no idea what this drags in

const { setGlobalDispatcher, Agent } = require("undici");

but that is why I linked to the solution ahead of time..

[bugsounet]

/!\ resolve and integrity are removed from package-lock.json

For information it's allows to download each libraries from an compressed file
Without it, MagicMiror deps will take many many Time for installing

[sdetweil]

@bugsounet I don't know what you are telling me.. git says there are two lines changed in package.json

systeminfo(changed version) and undici (added)

after I changed, I erased package-lock.json , and did a new npm install to get the latest

[khassel]

@sdetweil how did you test this? With an arm v6 machine or can I simulate this some other way?

[sdetweil]

i tested on the armv6l that had the timeout error on both news and calendar and also on my amd64 machine.

[khassel]

what url did you use for testing/simulating?

[sdetweil]

the default config urls. on armv6l
and my google cal and default on my amd

i had hard coded it in calendar to
prove, news failed

thats why the question about fixing for any/all via app/whatever
then put this on the armv6, and both worked
and of course all the testcases e2e and electron

[sdetweil]

only way to simulate would be to create some cpu sucker threads to cause delays . armv6 problem is small memory and slow processor handling swapping on sd card

[sdetweil]

i wonder if the guys 11 calendars failure would be fixed here
https://forum.magicmirror.builders/topic/19031/calendar-module-error-with-many-google-calendars?_=1734883587981

[khassel]

@sdetweil my question was only how I can test it, used this not working url now.

Your solution works, my tests with --network-family-autoselection-attempt-timeout does not work. The undici package is small, so I'm fine with merging this beside the package-lock.json problem.

If you look at the diffs of package-lock.json you removed many many lines like

			"resolved": "https://registry.npmjs.org/@altano/repository-tools/-/repository-tools-0.1.1.tgz",
			"integrity": "sha512-5vbUs2A98CC3g1AlOBdkBE0BMukkLjLIsMHAtuxg6Pt9dQXxYWdLKOf66v6c/vIqtNcgTMv0oGtddLdMuH9X6w==",

Don't know how you did it but this must be reverted before merge ...

[sdetweil]

@khassel oh, how to test timeout
use this

https://httpbin.org/delay/10

the number is seconds of delay

[sdetweil]

If you look at the diffs of package-lock.json you removed many many lines like

I did NOT "remove" anything

I see the resolved is missing on some, and the integrity value is different.. I just ran npm install with no package-lock, and no node_modules folder to generate a new one , this IS a different node version..

npm install (not install-mm) on my amd64 is blazing fast, <30 seconds total

[khassel]

I did NOT "remove" anything

look at the diff

[sdetweil]

@khassel I saw the diff.. but I didn't DO it on purpose.. just npm install

[khassel]

seems to be a npm issue, maybe this helps: npm/cli#4263 (comment)

[sdetweil]

I ran npm install again without removing the files and it looks better..
pushed

[bugsounet]

Remove node_helper and do npm install again
Diff package-lock.json and see result.
Must be fine

[sdetweil]

package-lock.json did not change

[bugsounet]

I can't verify from my computer, sorry

[sdetweil]

I can't verify from my computer, sorry

I don't know what you are telling me

[bugsounet]

@sdetweil : verified it's ok for me :)