Hi all,
I'm investigating an interesting meltdown of the folding system
when opening CGI.pm from Perl 5.8.2 with the recent Perl package
folding feature enabled.
Here is my current understanding of the failure:
Behaviour: When opening CGI.pm (this happens on opening only, I
think), the editor tries to fold an inordinately large length of
the file beyond the visible lines. The folding system goes through
several levels of nesting and too many EnsureStyledTo() calls,
leading to either the editor becoming unresponsive for a short
time or an outright crash.
When package folding is enabled, everything after a file-level
package statement is at level 1025 or above, while the package
statement itself is at 1024.
The failure is triggered when the last line of a folding request
is a fold header. SetLevel() triggers a notification, which ends
up in FoldChanged(). FoldChanged calls Expand(), ending up in
GetLastChild(). In GetLastChild(), an EnsureStyledTo() chain
reaction occurs.
In GetLastChild(), for the while loop, the IsSubordinate() test
always has 1024 as levelStart on the package statement line, and
everything after that will be >= 1025, so levelStart < levelTry
holds for the entire block and the while loop continues until the
end of the package block, spewing lots of EnsureStyledTo() calls.
There is also some nesting and redundant behaviour I don't fully
understand.
I'm getting something like this:
lineMaxSubord:1964,maxLine:7125
EnsureStyledTo(1966)
IsSubordinate(400,401)
lineMaxSubord:1965,maxLine:7125
EnsureStyledTo(1967)
IsSubordinate(400,401)
lineMaxSubord:1966,maxLine:7125
EnsureStyledTo(1968)
IsSubordinate(400,401)
lineMaxSubord:1967,maxLine:7125
EnsureStyledTo(1969)
IsSubordinate(400,401)
...
Still studying the problem, not sure what to do yet. I can see
that GetLastChild() does try to fold a whole child block, so maybe
a package block is too large or complex for the current folding
implementation to handle.
When the editor requests the whole file to be folded, there is no
such problem as I suppose EnsureStyledTo() does not cause extra
folding requests to be queued.
--
Cheers,
Kein-Hong Man (esq.)
Kuala Lumpur, Malaysia
_______________________________________________
Scintilla-interest mailing list
[email protected]
http://mailman.lyra.org/mailman/listinfo/scintilla-interest
