Please find logs and Video @ http://go/chrome-androidlogs1/6/691506

Good commit: 445242
Bad commit: 445243
Suspect CL: https://chromium.googlesource.com/chromium/src/+/fe02d6b0fd390b03345187b65e29831f4fb89ac7

@carlosk could you please look into this issue, thanks!

 Issue 687536  has been merged into this issue.

I started investigating this and it seems we're not getting neither DocumentAvailableInMainFrame or DocumentOnLoadCompletedInMainFrame signals from the WebContents. Without those the code assumes the page is loading forever and so the download waits indefinitely.

This doesn't seem to be related to my change.

This is my test procedure: on a new tab page I search for Tesla and click on one of the AMP links. From that point on this is what happens in terms of loading signals:

1) DidFinishNavigation is called for https://www.google.com/search?ei=djOiWNPQH82ijwO99quoDQ&q=tesla&oq=tesla&gs_l=mobile-gws-serp.3..0i67k1l2j0i131k1j0i67k1j0i131k1.28760.30279.0.31321.6.6.0.2.2.0.442.1612.0j2j2j1j1.6.0....0...1c.1j4.64.mobile-gws-serp..0.6.1326.2..0j41j0i131i67k1.ynRZRGgxGCA#scso=uid_WKI0ZgAMwy8KY-gTRAnvfA_0:0

2) DidFinishNavigation is called for https://www.google.com/amp/gizmodo.com/what-we-know-about-unionization-efforts-at-teslas-fremo-1792300563/amp

3) DidFinishNavigation is called on a subframe for https://www-marketwatch-com.cdn.ampproject.org/v/www.marketwatch.com/amp/story/guid/E6CA6E62-F220-11E6-82ED-7800910FCE87?amp_js_v=7#origin=https%3A%2F%2Fwww.google.com&exp=a4a%3A0&channelid=0&cid=1&dialog=0&prerenderSize=1&visibilityState=prerender&paddingTop=54&history=1&p2r=0&horizontalScrolling=0&csi=0&storage=1&viewerUrl=https%3A%2F%2Fwww.google.com%2Famp%2Fwww.marketwatch.com%2Famp%2Fstory%2Fguid%2FE6CA6E62-F220-11E6-82ED-7800910FCE87

But there's no calls at all to any of the two WebContentsObserver signals we rely on for knowing when the page is loaded: DocumentAvailableInMainFrame and DocumentOnLoadCompletedInMainFrame.

More findings:
* Other observer methods are are *not* being called *for the main frame*: DocumentLoadedInFrame and DidFinishLoad.
* None of the 4 referred methods are neither called when I hit the "back" button, when navigating back to the Google search results paged from the AMP page.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/f0b30fafe9d73f73c430895d289ab83c3e818d20

commit f0b30fafe9d73f73c430895d289ab83c3e818d20
Author: carlosk <carlosk@chromium.org>
Date: Tue Feb 21 19:37:00 2017

Last_n: create snapshots from user triggers even if similar quality ones exist.

With this change when a tab-hidden event happens a last_n snapshot will
be created even if one already exists with the same expected quality
level. This should fix cases of very dynamic pagesnot being properly
saved, where contents are completely changed without clear navigation or
loading events being triggered.

This also re-introduces the ignoring of same-page navigations when
listening to DidFinishNavigation so that we don't reset the loading
state of the page in those cases.

BUG= 691506 , 678367 

Review-Url: https://codereview.chromium.org/2698713002
Cr-Commit-Position: refs/heads/master@{#451819}

[modify] https://crrev.com/f0b30fafe9d73f73c430895d289ab83c3e818d20/chrome/browser/android/offline_pages/recent_tab_helper.cc
[modify] https://crrev.com/f0b30fafe9d73f73c430895d289ab83c3e818d20/chrome/browser/android/offline_pages/recent_tab_helper.h
[modify] https://crrev.com/f0b30fafe9d73f73c430895d289ab83c3e818d20/chrome/browser/android/offline_pages/recent_tab_helper_unittest.cc

Any reason why this issue should remain Googler-restricted?

This issue is fixed on latest M58-58.0.3021.0

 Issue 663290  has been merged into this issue.