cc dominick in case you have ideas why the pref provider might return two conflicting rules for the same origin

I've done some poking and my suspicion is because of the inconsistent treatment of the port between url::Origin::Serialize -> ContentSettingsPattern::FromString and GURL -> ContentSettingsPattern::FromURLNoWildcard.

url::Origin::Serialize() strips the port if it's the default port for the scheme (https://cs.chromium.org/chromium/src/url/scheme_host_port.cc?type=cs&sq=package:chromium&l=217). Then using ContentSettingsPattern::FromString sets the port to a wildcard  (https://cs.chromium.org/chromium/src/components/content_settings/core/common/content_settings_pattern_parser.cc?gsn=Parse&l=167).

However, ContentSettingsPattern::FromURLNoWildcard (which is used by SetWebsiteSetting) explicitly keeps the port, and if the port isn't there, it adds the default port, not a wildcard (https://cs.chromium.org/chromium/src/components/content_settings/core/common/content_settings_pattern.cc?type=cs&q=components/content_settings/core/common/content_settings_pattern.cc&sq=package:chromium&l=387). So you have this behaviour:

1. Allow permission (port = 443)
2. Migrate (creates channels with the port stripped out)
3. Unmigrate (uses ContentSettingsPattern::FromString on url::Origin::Serialize, no port)

At this point we have two conflicting rules: one for port 443, and one for *.

Can you try keeping to using the same code path as what settings does (i.e. ContentSettingsPattern::FromURLNoWildcard from a GURL)?

Agree with your hypothesis - adding a log in content_settings_pref_provider confirmed it:
... unmigrate ...
08-25 10:46:20.477 W/chromium(25834): [WARNING:content_settings_pref_provider.cc(149)] ANITA: SetWebsiteSetting pattern=https://tests.peter.sh

... block site from site settings ...
08-25 10:47:22.513 W/chromium(25834): [WARNING:content_settings_pref_provider.cc(149)] ANITA: SetWebsiteSetting pattern=https://tests.peter.sh:443


, changing ContentSettingsPattern::FromString(..) --> ContentSettingsPattern::FromURLNoWildcard(GURL(...)) where I construct the pattern from the channel origin means they both get the port and only one rule is then returned so the crash goes away - hooray! 

Thanks so much for your help figuring this out - saved me a bunch of debugging :)

Now I just need to figure out how to write a regression test for this!

I've done a quick fix in the interests of expediency as per #3 here: https://chromium-review.googlesource.com/c/chromium/src/+/643069 

Filed  crbug.com/760510  to make it use the same codepath as SetWebsiteSetting in future as suggested in #2. I think that's a really good idea but it'll take a little longer.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/42bb1dd7d8e7836997e0207d472e15c65bbc2aad

commit 42bb1dd7d8e7836997e0207d472e15c65bbc2aad
Author: Anita Woodruff <awdf@chromium.org>
Date: Wed Aug 30 14:11:52 2017

[Android notification channels] Quick fix for flag re-enabling crash

- Previously the ContentSettingsPrefProvider could end up storing two
conflicting notification settings for the same origin after
unmigrating channel settings, due to recreating the
ContentSettingsPattern from the channel origin inconsistently with how
it is created when setting a website pattern in the normal way. This
meant we would sometimes fail a DCHECK in the channels provider when
re-migrating to channel settings.

- This change makes the construction of a ContentSettingsPattern from
a channel origin string consistent with how ContentSettingPatterns are
constructed via HostContentSettingsMap.SetWebsiteSetting - first
convert to a GURL, then use ContentSettingsPattern::FromURLNoWildCard.

- A proper fix would involve going through HCSM.SetWebsiteSetting,
to ensure this is always done consistently - filed  crbug.com/760510 
to do this.

Bug:  758635 
Change-Id: I468075e62f7dce7f092389fb72f9d02f89925902
Reviewed-on: https://chromium-review.googlesource.com/643069
Commit-Queue: Anita Woodruff <awdf@chromium.org>
Reviewed-by: Peter Beverloo <peter@chromium.org>
Cr-Commit-Position: refs/heads/master@{#498451}
[modify] https://crrev.com/42bb1dd7d8e7836997e0207d472e15c65bbc2aad/chrome/browser/notifications/notification_channels_provider_android.cc
[modify] https://crrev.com/42bb1dd7d8e7836997e0207d472e15c65bbc2aad/chrome/browser/notifications/notification_channels_provider_android_unittest.cc