Related issues:
- Issue 655708: "Offline Page Cache: allow MHTML file save to be successful even if frames are missing"
-  Issue 655697 : "Offline Page Cache: do not delete previous save if last one fails"

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/29da4fe0a2b453e474e6cd530332a2ed02f70576

commit 29da4fe0a2b453e474e6cd530332a2ed02f70576
Author: carlosk <carlosk@chromium.org>
Date: Tue Nov 29 00:01:29 2016

Fail when saving page as MHTML provides information about the cause.

This change adds more detailed information to tracing and UMA histograms about
the reason why a MHTML save operation has failed. This will help us understand
why some specific pages consistently fail and better handle these situations.

It introduces a new enum, MhtmlSaveStatus, that lists all tracked causes. Part
of the reasons are tracked by the browser and the others by the render process
but the reporting is centralized in the former. A few tests from
MHTMLGenerationTest were also changed to verify the MHTML failures they test for
are correctly reported.

BUG= 645686 , 655723 ,655708
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_site_isolation

Review-Url: https://codereview.chromium.org/2519273002
Cr-Commit-Position: refs/heads/master@{#434783}

[modify] https://crrev.com/29da4fe0a2b453e474e6cd530332a2ed02f70576/content/browser/download/mhtml_generation_browsertest.cc
[modify] https://crrev.com/29da4fe0a2b453e474e6cd530332a2ed02f70576/content/browser/download/mhtml_generation_manager.cc
[modify] https://crrev.com/29da4fe0a2b453e474e6cd530332a2ed02f70576/content/browser/download/mhtml_generation_manager.h
[modify] https://crrev.com/29da4fe0a2b453e474e6cd530332a2ed02f70576/content/browser/frame_host/render_frame_host_impl.cc
[modify] https://crrev.com/29da4fe0a2b453e474e6cd530332a2ed02f70576/content/browser/frame_host/render_frame_host_impl.h
[modify] https://crrev.com/29da4fe0a2b453e474e6cd530332a2ed02f70576/content/common/BUILD.gn
[modify] https://crrev.com/29da4fe0a2b453e474e6cd530332a2ed02f70576/content/common/content_param_traits_macros.h
[add] https://crrev.com/29da4fe0a2b453e474e6cd530332a2ed02f70576/content/common/download/mhtml_save_status.cc
[add] https://crrev.com/29da4fe0a2b453e474e6cd530332a2ed02f70576/content/common/download/mhtml_save_status.h
[modify] https://crrev.com/29da4fe0a2b453e474e6cd530332a2ed02f70576/content/common/frame_messages.h
[modify] https://crrev.com/29da4fe0a2b453e474e6cd530332a2ed02f70576/content/renderer/render_frame_impl.cc
[modify] https://crrev.com/29da4fe0a2b453e474e6cd530332a2ed02f70576/content/renderer/render_frame_impl.h
[modify] https://crrev.com/29da4fe0a2b453e474e6cd530332a2ed02f70576/tools/metrics/histograms/histograms.xml

Running a few tests using the new tracing information I confirmed our theory that the failures while offlining TechCrunch are due to missing iframes. The work on allowing MHTML generation to succeed when frames are missing should help or even fully resolve this error.

The OP suggested investigation is done and issue 655708 was created to solve the detected "frames disappearing" issue. So I'll mark this one as fixed.