------- You are receiving this mail because: -------
You are on the CC list for the bug.
http://bugs.exim.org/show_bug.cgi?id=1370
Summary: Combined graphemes in UTF8 may cause segmentation fault
Product: PCRE
Version: 8.33
Platform: x86-64
OS/Version: Linux
Status: NEW
Severity: bug
Priority: high
Component: Code
AssignedTo: ph10@???
ReportedBy: bufflig@???
QAContact: bufflig@???
CC: pcre-dev@???
Created an attachment (id=642)
--> (
http://bugs.exim.org/attachment.cgi?id=642)
Program to provoke the crash and dirty fix diff
The PCRE library (8.33) was built with
$ ./configure --enable-utf --enable-unicode-properties --disable-shared
--disable-stack-for-recursion
$ make
A program with a subject containing combined graphemes (like an 'A' followed by
a combining diacritical mark, UCD_CATEGORY ucp_M) may cause the eptr to back of
to positions prior to the actual subject and eventually cause a segmentation
fault.
The problem *shows up* in pcre_exec.c:5684 and the loop below. A quick fix
returning MATCH_NOMATCH if eptr moves back before md->start_subject removes the
crash, but I am not sure if this is the semantically correct fix.
I've attached a tar.gz file with a diff for my dirty fix, just to point out
where the error occurs, as well as a test program (a stripped down version of
pcredemo.c, which places some garbage before the actual subject to provoke the
crash). The test program dumps core on my 64bit Linux.
I found the error while integrating PCRE 8.33 into the next version of
Erlang/OTP, as I by mistake run test number 10 on the regular machine instead
of the DFA ditto.
Please see attached files for details.
Best regards,
Patrik Nyblom, Erlang/OTP
--
Configure bugmail:
http://bugs.exim.org/userprefs.cgi?tab=email
