Redirects in main document are not detected as new page (explicit redirect or navigation)

[xavier-medioni]

Hi everybody !

At first, thanks for this very useful tool !

And I wanted to suggest an improvement about a multi page HAR output, and I will try to provide as many elements as possible.

Usecase:

page1.html: loading 3 images + 1 js file, redirecting to page2.html on DOMContentloaded + 500ms page2.html: loading 2 images

With chrome-har:

• the generated HAR object is "flat" (only page1.html detected, and every entries are relative to this page => 1 page in har.log.pages)

With Chrome Devtools > Save all as HAR with content:

• the generated HAR object detects the 2 pages (page1.html and page2.html exist in har.log.pages, with entry.pageref relative to expected page id)

Suspicion:

• in chrome-har/index.js, method harFromMessages: add a page (maybe on "Page.frameScheduledNavigation" protocol event) when frameId == rootFrame ?

Alternative / extended behaviour:

• possibility to add an extra "allowMultiPage" parameter to harFromMessages to generate multi page HAR when set to true (preserve default behaviour) ?

Thanks by advance !

[soulgalore]

Hi @xavier-medioni thanks for creating the issue! So for redirects its been that by choice, with the idea that you choose to test one page, it redirects to the "same" page that you wanted to test.

For multi, yeah, I started to do that before but it started to get too complicated at that time. That's low prio for me but if someone is interested maybe I can guide you through it.

[xavier-medioni]

Hi @soulgalore,

Thank you for your reply. As you probably already guessed, I was trying to evaluate, from a network point of view, the cost of a potential redirect, capturing protocol events "live", and using harFromMessages. I tried to tweak the existing harFromMessages method to add page on "Page.frameScheduledNavigation" event (maybe the most convenient event to track to detect redirects), but got stucked due to an exception because const entrySecs = page.wallTime + (timing.requestTime - page.timestamp); with page.wallTime and page.timestamp were null. I noticed the use of addFromFirstRequest, (addFromFirstRequest(page, paramsWithoutPage[0]), also called into "Network.requestWillBeSent" case), but not yet sure about how to use it for this usecase.

If you have any lead, I could be working on it.