Tracking Fission DevTools bugs for Fission Nightly (M6)

dt-fission-reserve bugs do not need to block Fission Nightly (M6).

Daisuke migrated to StyleSheet editor to the ResouceWatcher API on the client, making it Fission compatible.
But we are now hitting this bug, which prevents displaying stylesheets correctly when the page is already loaded:
https://phabricator.services.mozilla.com/D78240#2505525

This issue may actually prevent us from landing server side network request listening, i.e. bug 1644191.
So bumping this into the MVP list.

Tracking dt-fission-m2-mvp bugs for Fission Nightly milestone (M6c).

(In reply to Alexandre Poirot [:ochameau] from comment #4)

This issue may actually prevent us from landing server side network request listening, i.e. bug 1625930.

Alex, are you pointing to the right bug here?

Honza

I meant bug 1644191. I updated my older comment.

Hello, Alex!
If the reason why we needed to see consoleActor.getRequestContentForURL(href) was to use the cache in the console actor, as DevTools has become to fetch all contents from our cache in the platform (to fix other issues), we may not need to see the console actor anymore?

This has been introduced in bug 1306892.
It is about preventing re-fetching the stylesheets.
Now, I don't know if the cache is populated as much as the netmonitor?
If the cache is complete as netmonitor inspection, then yes, it sounds like a good fix!

I have some doubt when I read Patrick's comment (bug 1306892 comment 11):

What I do remember is that if a stylesheet is set not to be cached, then the back-end has no other choices then to actually download it again from the server.
Could it be that some HTTP header will prevent the stylesheet from being cached, and we will redo the request here?

Thank you very much for your feedback, Alex!
Okay, let me check again whether the cache covers with all of the contents that the document loaded or not.

I have investigated with ignoring the result of consoleActor.getRequestContentForURL that where the stylesheet that DevTools got is taken from.

• debug patch: https://hg.mozilla.org/try/rev/42049ad85065c868f6501908da3e51648d121757
• try: https://treeherder.mozilla.org/#/jobs?repo=try&selectedTaskRun=Sw4PUVm6QnWlA2wE7z_kpQ.0&revision=6e036e538488b6f4fd3f8c3b514a324793efe2a6
• log that includes styleeditor tests: https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=308778961&repo=try&lineNumber=8536

As a result, we found that most of stylesheets are taken from the cache in the platform expectedly.
e.g.
https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=308778961&repo=try&lineNumber=8538

One thing we could not take from the cache was imported CSS.
https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=308778961&repo=try&lineNumber=8691
But, as it seems we can get from the cache when browsing the page normally, it might be caused by the timing?

On the other hand, tests that tried to use consoleActor are browser_styleeditor_navigate.js and browser_styleeditor_fetch-from-netmonitor.js.
For the browser_styleeditor_navigate.js as well, we can confirm that the stylesheet can be taken from the cache in the platform.
https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=308778961&repo=try&lineNumber=8994
For the browser_styleeditor_fetch-from-netmonitor.js, as this test checks whether we can get the proper content from consoleActor, it will be failed here, but can ignore.
https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=308778961&repo=try&lineNumber=8617-8622

As a result, there will be no problem if we get rid of referring to consoleActor I think, but what do you think?
And, if you have any better ways that we can check whether the stylesheets are taken from the cache or not, please let me know!

(In reply to Alexandre Poirot [:ochameau] from comment #9)

I have some doubt when I read Patrick's comment (bug 1306892 comment 11):

What I do remember is that if a stylesheet is set not to be cached, then the back-end has no other choices then to actually download it again from the server.
Could it be that some HTTP header will prevent the stylesheet from being cached, and we will redo the request here?

I think this was fixed.
For example, even if set disable cache checkbox in the network monitor, DevTools can get from the cache (BF cache?) in the platform in order to get exactly the same content.

I heard from Honza on slack, as Alex and Honza agreed with using the browser cache, I will try to fix it with that way.
Thanks!

Depends on D83287

Depends on D83288

https://hg.mozilla.org/mozilla-central/rev/e16c7d81d82a
https://hg.mozilla.org/mozilla-central/rev/21078a5722b9
https://hg.mozilla.org/mozilla-central/rev/e81d80110147

Depends on D85370

https://hg.mozilla.org/mozilla-central/rev/153dc423b227
https://hg.mozilla.org/mozilla-central/rev/11e97601cbca