Incorrect regex matching

[Zantier]

Version 6.7.7.
I've removed all plugins apart from the plugin manager.

Here is the simplest case I've came up with:

• Open a new tab
• Type/paste at least 380 "a"s
• Ctrl+F -> with "Regular expression" search mode selected, search for ([^b])*b

1 occurrence is found, which matches the whole text, when of course the document doesn't contain "b".

[tommilligan]

Interesting bug. Replicated easily (v6.7.7). Confirmed that only happens after exactly 380 characters. Bug not present if () removed; (a)*b gives bug but a*b does not even after 200,000 characters.

Suspect memory/storage issue of () capturing group?

[ghost]

Oddly enough I get the bug after 270 characters.

[Zantier]

It appears to be an issue with Scintilla, as I was able to reproduce a similar behaviour in SciTE by dropping in SciLexer.dll v3.3.4 (which Notepad++ v6.7.7 is using) in place of v3.5.5.

When I place SciLexer.dll v3.5.5 in the same directory as Notepad++, I no longer see the issue, even with 200,000 "a"s in a row. Not that SciLexer.dll v3.5.5 is perfect though, as it seems to fail to match (a)*, or a more useful regex such as (a|b)* in both Notepad++ and SciTE...

---

Edit: Note I don't know whether SciLexer.dll v3.5.5 was built with boost for PCRE support. I simply took SciLexer.dll from the SciTE binary download. I don't have time to check right now.

[guy038]

Hello Zantier, tommilligan, jonandr and All,

I did some tests, on a 6.7.9 version, with its native plugins, and I found out a strange rule about that regex issue ! Follow the few steps below, to reproduce it :

• Let's suppose that you write, in a new tab, a line with several lower letters a ( let's say, for instance, between 10 and 20 )
• Then, just add some lines, even empty, before and after this line of a's letters, with the two following conditions :

1. They don't contain any lower letter a or b

2. If your select all the contents of that test file, you get, exactly, 285 characters

• Go back to the very beginning of the file ( CTRL + Origin )
• Open the Find dialog
• Check the Match case option and the Regular expression radio button
• Uncheck the option Wrap around
• Type the Zantier's regex ([^b])*b in the Find what zone

=> When clicking on the Find Next button, you wrongly get the entire contents of this test file :-(

Now, :

• Close the Find dialog , hitting the ESC key
• Delete ONE character ONLY, in a line, located before, or even after, the line of a's ( so, if you hit CTRL-A, you get, exactly, a 284 characters selection )
• Go back, again, to the very beginning of the file, with the CTRL + Origin shortcut
• Type on the F3 key to repeat the same regex search ([^b])*b

=> This time, the regex engine, doesn't find, as expected, any matched string ( as the file contains NO lower letter b )

So, for some odd reasons, this regex doesn't work, as soon as the size of the file is > 284 bytes ???

Then, I replace, successively, the SciLexer.dll v3.5.6 file by :

• The last version of SciLexer.dll v3.3.4 ( which came with N++ 6.7.3 ) => Same limit, between 284 and 285
• The last version of SciLexer.dll v2.2.7 ( which came with N++ 6.3.0 ) => The limit is between 269 and 270

---

BTW, the regex ([^b])+b produces the same issue, but, luckily, the regexes ([^b]*)b ( that, however, creates a different group 1 ) and the regex [^b]*b, do work, in all cases, independently of the file"'s size !

Best regards,

guy038

[robert-andrzejuk]

Has anyone reported this on the Scintilla bug system?

[donho]

It's an issue concerning Scintilla.

[sasumner]

It's an issue concerning Scintilla

The symptoms sound like Notepad++ isn't catching an exception that it should (see #4761 (comment)).

Anyway, if it is really thought to be a Scintilla issue, why not update Scintilla and see if it remains. :-)