Problem: Matching with a virtual column has a lot of overhead on very long
lines. (Issue 310)
Solution: Bail out early if there can't be a match. (Christian Brabandt)
Also check for CTRL-C at every position.
Problem: With some regexp patterns the NFA engine uses many states and
becomes very slow. To the user it looks like Vim freezes.
Solution: When the number of states reaches a limit fall back to the old
engine. (Christian Brabandt)
Problem: A non-greedy match followed by a branch is too greedy. (Ingo
Karkat)
Solution: Add NFA_MATCH when it is already in the state list if the position
differs.
Problem: Pattern with repeated backreference does not match with new regexp
engine. (Urtica Dioica)
Solution: Also check the end of a submatch when deciding to put a state in
the state list.
Problem: Crash when using cpp syntax file with pattern using external
match. (Havard Garnes)
Solution: Discard match when end column is before start column.
Problem: NFA regexp doesn't handle backreference correctly. (Ryuichi
Hayashida, Urtica Dioica)
Solution: Always add NFA_SKIP, also when it already exists at the start
position.
Problem: Syntax highlighting a Yaml file causes a crash. (Blake Preston)
Solution: Copy the pim structure before calling addstate() to avoid it
becoming invalide when the state list is reallocated.
Problem: Using "\ze" in a sub-pattern does not result in the end of the
match to be set. (Axel Bender)
Solution: Copy the end of match position when a recursive match was
successful.
Problem: Equivalence classes are not working for multi-byte characters.
Solution: Copy the rules from the old to the new regexp engine. Add a test
to check both engines.
Problem: NFA regexp: Using \ze in one branch which doesn't match may cause
end of another branch to be wrong. (William Fugh)
Solution: Set end position if it wasn't set yet.
Problem: Pattern with two alternative look-behind matches does not match.
(Amadeus Demarzi)
Solution: When comparing PIMs also compare their state ID to see if they are
different.
Problem: Character classes such as [a-z] to not react to 'ignorecase'.
Breaks man page highlighting. (Mario Grgic)
Solution: Add separate items for classes that react to 'ignorecase'. Clean
up logic handling character classes. Add more tests.
Problem: Crash searching for \%(\%(\|\d\|-\|\.\)*\|\*\). (Marcin
Szamotulski) Also for \(\)*.
Solution: Do add a state for opening parenthesis, so that we can check if it
was added before at the same position.
Problem: New regexp engine doesn't match pattern. (Ingo Karkat)
Solution: When adding a state also check for different PIM if the list of
states has any state with a PIM.
Problem: Reading memory already freed since patch 7.3.1247. (Simon
Ruderich, Dominique Pelle)
Solution: Copy submatches before reallocating the state list.
Problem: New regexp engine: '[ ]\@!\p\%([ ]\@!\p\)*:' does not always match.
Solution: When there is a PIM add a duplicate state that starts at another
position.
