OverflowError: regular expression code size limit exceed

[Ryan J Ollos]

It was reported in ​trac:#10190 that the regular expression can become too large on a site with many wiki page.

A similar issue was also reported this week for the AcronymsPlugin in a private email exchange. The two plugins are very similar, so if we resolve the issue for one, we may have insight into solving the issue for the other.

[Chris Nelson]

What I find odd is that this broke on our upgrade from Trac 0.11.6 to 1.0.1. I believe our version of Python and this plugin remained the same.

[Chris Nelson]

Replying to ChrisNelson:

What I find odd is that this broke on our upgrade from Trac 0.11.6 to 1.0.1. I believe our version of Python and this plugin remained the same.

BTW, this works in my sandbox but fails in production. Both are running the current 0.2dev Acronyms plugin, Trac 1.0.1 and the same version of Python. I'm trying to identify differences.

[Jun Omae]

I tried to reproduce this issue on Trac 1.0.1 with AutoWikifyPlugin and AcronymsPlugin. However, I cannot at the moment. I've created 83,000 wiki pages and installed the same of AcronymDefinitions in trac-hacks.org.... Could someone provide how to reproduce?

[Chris Nelson]

I now suspect it's a plugin interaction. Our production has more plugins installed than my sandbox. I'm going to work on closing that gap shortly.

[Chris Nelson]

I'm more sure of a plugin interaction, or at least pile up. When I do:

sre_compile.py

@@ -504 +504 @@
     code = _code(p, flags)
 
     # print code
+    file = open("/tmp/re.txt", "a");
+    file.write("re groups = %d\n" % p.pattern.groups)
+    file.close
 
     # XXX: <fl> get rid of this limitation!
     if p.pattern.groups > 100:

I have 98 groups with Acronyms Plugin enabled and 95 without so Acronyms is only adding three. In my production environment where this fails, I have 98 without, 101 with, thus crossing the threshold.

[Jun Omae]

I have 98 groups with Acronyms Plugin enabled and 95 without so Acronyms is only adding three. In my production environment where this fails, I have 98 without, 101 with, thus crossing the threshold.

In that case, I guess that AssertionError: sorry, but this version only supports 100 named groups is raised, not OverflowError.

>>> import re
>>> re.compile('|'.join('(?P<i%d>i)' % _ for _ in xrange(99)))
<_sre.SRE_Pattern object at 0x00B4A148>
>>> re.compile('|'.join('(?P<i%d>i)' % _ for _ in xrange(100)))
Traceback (most recent call last):
  File "<stdin>", line 1, in ?
  File "C:\usr\apps\python24\lib\sre.py", line 180, in compile
    return _compile(pattern, flags)
  File "C:\usr\apps\python24\lib\sre.py", line 225, in _compile
    p = sre_compile.compile(pattern, flags)
  File "C:\usr\apps\python24\lib\sre_compile.py", line 506, in compile
    raise AssertionError(
AssertionError: sorry, but this version only supports 100 named groups

[Chris Nelson]

Replying to jun66j5:

I have 98 groups with Acronyms Plugin enabled and 95 without so Acronyms is only adding three. In my production environment where this fails, I have 98 without, 101 with, thus crossing the threshold.

In that case, I guess that AssertionError: sorry, but this version only supports 100 named groups is raised, not OverflowError. ...

Yes, you are right. My issue appears to be different from the subject of this ticket.