Some rules with $document modifier work incorrectly

[AdamWr]

I'm not sure if it's a bug or just CoreLibs limitation.

Steps to reproduce

1. Add this rule:
   ||reddit.*^$document,domain=~reddit.com
2. Go here - https://duckduckgo.com/?q=site%3Areddit.com&ia=web
3. Open/Click on any link

Expected behavior

Website shouldn't be blocked.

Actual behavior

Website is blocked and if I refresh website then it's not blocked anymore.

Screenshot:

It's also not blocked if I open it directly - https://www.reddit.com/

Your environment

• AdGuard for Windows version 7.6 nightly 22 (build 3500, CL 1.7.202)

[sxgunchenko]

Hm... In which browser do you check it?
It seems like firefox and chrome act differently: FF sets the referrer to https://reddit.com/, but chrome sets it to https://duckduckgo.com/.
As a result, I can reproduce it in chrome, but cannot in FF.
In both cases the behavior is correct.

[AdamWr]

In which browser do you check it?

In Chrome.

Screenshot

If request URL is excluded in a rule (I mean domain=~reddit.com), shouldn't it be unblocked anyway?
If not, then as far as I understand it's CoreLibs limitation, with Chrome extension it works fine and website is not blocked.

[sxgunchenko]

The logic of matching rules with the $domain modifier was slightly different in CL 1.7.202. A request with the document resource type matches against a rule with the $domain modifier if its referrer or target host matches against the modifier.

In this case, the referrer (duckduckgo.com) matches against the modifier (~reddit.com).

UPD: as for exclusions in the $domain modifier, a rule does not match if the referrer is excluded, but not the target domain

[sxgunchenko]

See #1354 (comment) for the upcoming changes

[sxgunchenko]

Yup, it seems like I need the example.

Tried some combinations of target and referrer and it works as expected (target = t, referrer = r) :
t = http://advertisement.com/some, r = http://example.com => blocked as rule's pattern matches the target
t = http://advertisement.com/some, r = http://advertisement.solutions.zalando.com => not matched as the referrer is explicitly excluded
t = http://advertisement.solutions.zalando.com/some, r = http://advertisement.com => blocked as the pattern matches the target (see my previous comment #1421 (comment) about target exclusion)

[krystian3w]

Maybe a little help: https://kb.adguard.com/en/general/how-to-create-your-own-ad-filters#removeheader

[AdamWr]

It seems that now, rule like this:

||reddit.*^$document,domain=~reddit.com

blocks reddit.com, but I think that it shouldn't be blocked.

Steps to reproduce

1. Add this rule:

||reddit.*^$document,domain=~reddit.com

2. Go to - https://www.reddit.com/

Screenshot

reddit.com shouldn't be blocked.

AdGuard for Windows version 7.9 nightly 3 (build 3789, CL 1.8.279)

[krystian3w]

Also uBo started mark *^ in that filters as no needed due ^ no protect(/exclude scan) path after TLD/ or TLD./.

[Alex-302]

@krystian3w *^ parsed as TLD only?

[krystian3w]

uBo use gray color in highlighter, so filter writters can remove this form projects.

semi Hi-DPI screenshot

IMO nice will be implement enforce limit to tld for .*^ instead only mark as no power syntax - now may possible by regex only in uBo:

/^https?:\/\/([\w]{1,4}\.)?reddit\.[\w]{2,14}/$all,domain=~reddit.com

But page may use many subdomains like:

very.bad.gaga.reddit.angency to omit my regex or tld with 15 characters (maybe possible add safe . to allowed searched charactes without high CPU useage).

[sxgunchenko]

It seems that now, rule like this:

||reddit.*^$document,domain=~reddit.com

blocks reddit.com, but I think that it shouldn't be blocked.

Steps to reproduce

1. Add this rule:

||reddit.*^$document,domain=~reddit.com

2. Go to - https://www.reddit.com/

Screenshot

reddit.com shouldn't be blocked.

AdGuard for Windows version 7.9 nightly 3 (build 3789, CL 1.8.279)

Moved into a separate issue: #1579