Logged In: YES 
user_id=86307

Gustavo, in my understanding, it doesn't matter if the first part 
of MAX_UNTIL is protected because if its call to 
SRE_MATCH returns 0, it (the first part of MAX_UNTIL) 
simply returns 0 to its caller.  If that caller is a backtrack 
point (a point which may resume matching even though a call 
to SRE_MATCH returned 0), it should have already saved 
lastmark, etc. and so should be ready to restore them to its 
saved values.  If that caller is not a backtrack point, it is 
either the original invocation of SRE_MATCH (in which case it 
doesn't matter what is in the state since the match has failed) 
or it is someplace which is going to return 0 to its caller.

FYI, I have uploaded a couple of more sre bug 
reports/patches in the last couple of days; I'd appreciate it if 
you look at them.

    http://www.python.org/sf/725106
    http://www.python.org/sf/725149

In particular, the second patch changes ASSERT_NOT so 
that (I hope) it no longer violates the assumptions above.