GitLab

In the current implementation, sometimes the scope of a page's "pause"
is too wide compared to the definition in the spec [1], and sometimes it
does not pause at all. So narrow the scope to conform to the spec, and
apply it consistently to all possible ways users can print.

This is based on an earlier CL [2] which only did this for printing with
the system dialog, but not for Print Preview.

The first attempt at landing this CL [3] got reverted [4] because it
caused bf_cache_browser_tests failures. These tests did not run by
default as part of the CQ, which is how [3] passed the CQ. To fix this
issue, CL [5] landed to make the test less flaky.

Another bug [6] uncovered by the original CL [3] is a bit harder to
solve. The issue here is the case where PrintRenderFrameHelper tries to
unpause during blink::WebLocalFrameImpl teardown, and it is happening
too late. To address this, introduce blink::WebPrintClient, so the
WebLocalFrameImpl can tell the WebPrintClient (PrintRenderFrameHelper,
really) that it will be destroyed. Then PrintRenderFrameHelper can
unpause at the correct time.

[1] https://html.spec.whatwg.org/C/#printing-steps
[2] https://crrev.com/c/1669250
[3] https://crrev.com/797120
[4] https://crrev.com/797234
[5] https://crrev.com/799501
[6] https://crbug.com/1115475

CQ_INCLUDE_TRYBOTS=luci.chromium.try:android-bfcache-rel
CQ_INCLUDE_TRYBOTS=luci.chromium.try:linux-bfcache-rel

Bug: 956832
Change-Id: Ie328d90094f1b282126d51425200f8810819c3c6
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2353886Reviewed-by: Calder Kitagawa <ckitagawa@chromium.org>
Reviewed-by: Kent Tamura <tkent@chromium.org>
Reviewed-by: Yuki Shiino <yukishiino@chromium.org>
Commit-Queue: Lei Zhang <thestig@chromium.org>
Cr-Commit-Position: refs/heads/master@{#799772}