Network filter toString fix

[chrmod]

I've run into a weird edge case. NetworkFilter#toString() adds a right anchor to filters ending with a separator.

  1) Network filters
       toString
         pprint anchored hostnames:

      AssertionError: expected '||foo.com^|' to equal '||foo.com^'

This may potentially be correct, but some other part of our toolchain (abp2dnr) does not handle the right anchors and thus ends up with a broken filter (as DNR does not support right anchor either, afaict).

Anyways, resulting syntax ^| is extremely rare, I found only one filter using it in both uAssets and Easylist.

Can we make this change?

[remusao]

I think it's this case:

adblocker/packages/adblocker/src/filters/network.ts

Line 748 in 2fb07f0

// If the only symbol remaining for the selector is '^' then ignore it

Can you share an example of filter where you encountered the issue?

[remusao]

I think the semantics of this is a bit different now. Although I think it might not be 100% correct in the previous case either. I vaguely remember that this case was a little bit hacky.

[seia-soto]

uBlock Origin team replied me with a clarification: | is used to match start/end of address.

||hostname^|  => http://example.com/ & not http://example.com/blablabla

Many thanks to @ ghajini from uBO.

[remusao]

If I remember correctly this combination of flags was used in order to implicitly detect cases like ||foo.com^ (hostname match with no selector) and avoid having to create a specific flag for this. Then we detect and adapt matching logic here:

adblocker/packages/adblocker/src/filters/network.ts

Line 1655 in 2fb07f0

if (filter.hasFilter() === false) {

Maybe the better way would be to introduce a flag for that and remove the hack. It would also allow to have a 100% correct toString() implementation. But I think your tweak also makes sense as long as it does not break other cases you care about.

[seia-soto]

I'm leaving a note during investigation on ^| syntax. At the first time, I thought this should be considered as an invalid filter generally.

Also:

Search result on uBO/EL

However, there's actually a use of ^| syntax which is not documented.

• https://github.com/gorhill/uBlock/wiki/Static-filter-syntax#csp

[seia-soto]

If we are going to process | and ^| differently, we also need to fix the following case: '||foo.com^$3p' resolves to '||foo.com^|$3p'.

seia-soto@f82f0a1

Also, see #3681 (comment)