Reported internally.
The issue involves `removeparam` filters with a regex value. When
such filter was visited before a selfie was created, this would cause
the created selfie to persist a RegExp object, which can't be
serialized. This would cause exceptions to be thrown when uBO would
be subsequently loaded with the tainted selfie, since uBO would try
to execute a plain Object as a RegExp.
This will allow to find out what is specifically blocked on the
page reported as having issues, potentially saving time
when volunteers try to diagnose issues.
Related issue:
- https://github.com/uBlockOrigin/uBlock-issues/issues/1912
Related commit:
- 7713597e3e
In above related commit, uBO was modified to cache the URL of the
last clicked link, and to use this cached URL to test for the
legitimacy of the newly opened tab.
The current commit add back a test against the URL of the last
clicked link to avoid false positives when navigating from within
the newly opened tab.
Related discussion:
- a0a9497b4a (commitcomment-62560291)
The new setting, when disabled (enabled by default), allows a user
to prevent uBO from waiting for all filter lists to be loaded
before allowing network activity at launch. The setting is enabled
by default, meaning uBO waits for all filter lists to be loaded in
memory before unsuspending network activity. Some users may find
this behavior undesirable, hence the new setting.
This gives the option to potentially speed up page load at launch,
at the cost of potentially not properly filtering network requests
as per filter lists/rules.
For platforms not supporting the suspension of network activity,
the setting will merely prevent whatever mechanism exists on the
platform to mitigate improper filtering of network requests at
launch. For example, in Chromium-based browsers, unchecking the
new setting will prevent the browser from re-loading tabs for
which there was network activity while in "suspended" state at
launch.
As the trie is not immediately created, in order to speed up
launch time, the `domain=` option was stored in the filterRefs
array until it was moved to the trie.
This commit instead stores the `domain=` option into the trie
container's character buffer.
This commit will force-reload active tabs at launch for
environments not supporting suspend network request listeners,
or configured to not suspend network request listeners.
Refactored heuristics to collate set of origin-related
filter units are collated into a hostname trie, and
for better reuse of existing classes.
Generalized pre-test idea for bucket of filters, such
that in addition to origin-related filter units, there is
now a class to collate regex-based pattern-related units
into a new pre-test bucket class, FilterBucketIfRegexHits,
in order to test with a single regex test whether there is
a chance of a hit in the underlying bucket of filters.
Instances of these are rare, but at time of commit I found
this occurs with AdGuard France filter list.
Fine-tuned the "SNFE: Dump" output -- this new ability to
see the internal details of the SNFE has been really key
into finding/fixing issues during refactoring.
As was done with generic pattern-based filters, the source
string of regex-based filters is now stored into the
bidi-trie (pattern) buffer.
Additionally, added a new "dev tools" page to more
conveniently peer into uBO's internals at run time, without
having to do so from the browser's dev console -- something
which has become more difficult with the use of JS modules.
The new page can be launched from the Support pane through
the "More" button in the troubleshooting section.
The benchmark button in the About pane has been moved to this
new "dev tools" page.
The new "dev tools" page is for development purpose only,
do not open issues about it.
There are currently over 160 patterns with such pointless
trailing `*^` in uBO's filter lists, which ended up being
compiled as generic pattern filters (i.e. regex-based
internally), while the trailing `*^` accomplishes nothing
since it will always match the end of a URL ( `^` can
also match the end of URL).
This commit discards pointless trailing `*^` in patterns,
thus allowing most of those filters to be compiled as
plain pattern filters.
The syntax highlighter will reflect that a trailing
`*^` is pointless.
Rearrange logic to instantiate and add `important` filters
to the block realm when compiled lists are loaded instead
of when lists are compiled.
Additionally, removed now unused properties following
commit 68e14793cc.
Related issue:
- https://github.com/uBlockOrigin/uBlock-issues/issues/1863
As per internal discussion with team, best to have a simpler
scriplet, and which is hard-coded to work only on a specific
set of domains -- only those seen used by BAB.
Turns out the various benchmarks show no benefits when compiling
filters whose pattern contains a single wildcard character into
specialized classes which threat the pattern as two sub-patterns,
and actually there is a slight improvement in performance as per
benchamrks when treating these patterns as generic ones.
This also fixes the following related issue:
- https://github.com/uBlockOrigin/uBlock-issues/issues/1207
Fixed serious regression in previous dev build in applying
`csp=` filters. Reported internally by uBO team.
Promote usage of `removeparam` in code instead of `queryprune`,
which is to be deprecated.
Removed test against previously tested hostname in
FilterHostnameDict since as per various benchmark, the
test does not really help.
Remove serialization API in Node.js code as the API is now
present in SNFE itself.
All the auxiliary data structures must be fully loaded before
the data structure used as entry point is populated. The race
condition could lead to a case of the entry point data structure
being populated while the auxiliary data structures are still
unpopulated, potentially causing exceptions to be thrown at
launch when the static network filtering engine is queried.
I haven't been able to reproduce such exceptions -- but it
could happen on browsers which do not support being suspended
at launch time (i.e. chromium-based browsers).
Additionally, added convenience methods to easily
serialize/unserialize when SNFE is used as a npm package.
