Explaining the File:/// bug. See also http://openradar.appspot.com/13128709

disassembly.txt

In DDResultCopyExtractedURL in the DataDetectorsCore.framework, file:// URLs are sanity-checked with an assert:

0xCB86 loc_CB86:
0xCB86  lea     rsi, cfstr_File ; "file://"
0xCB8D  mov     rdi, rbx
0xCB90  call    _CFStringHasPrefix ; Check if the string starts with 'file://'
                                   ; Yes, this is case sensitive, which is why the test fails
0xCB95  test    al, al
0xCB97  jne     short loc_CBD4     ; If CFStringHasPrefix returns true, jump past the assert
                                   ; Otherwise, the following code triggers an assert:
0xCB99  lea     rdi, aCfstringhaspre ; "CFStringHasPrefix(urlVal, CFSTR(\"file:/"...
0xCBA0  lea     rsi, aSourcecache_51 ; "/SourceCache/DataDetectorsCore/DataDete"...
0xCBA7  lea     rdx, aDdresultcopy_1 ; "DDResultCopyExtractedURL"
0xCBAE  lea     r14, cfstr_WrongExtractio ; "wrong extraction: %@"
0xCBB5  mov     ecx, 628h
0xCBBA  mov     r8, r14
0xCBBD  mov     r9, rbx
0xCBC0  xor     al, al
0xCBC2  call    _DDLogAssertionFailure
0xCBC7  mov     rdi, r14
0xCBCA  mov     rsi, rbx
0xCBCD  xor     al, al
0xCBCF  call    _DDCrash

evil-fix.txt

Evil fix for x86_64 / 10.8.2. ____Don't actually use this____. This is just for entertainment.

We just change the 'jne' to non-conditional 'jmp':

printf '\xeb' | dd bs=1 seek=646039 count=1 conv=notrunc of=/System/Library/PrivateFrameworks/DataDetectorsCore.framework/Versions/Current/DataDetectorsCore

eg, 

0xCB97 jne     short loc_CBD4     ; If CFStringHasPrefix returns true, jump past the assert
.... becomes:
0xCB97 jmp     short loc_CBD4     ; Always jump past the assert

And now you no longer crash.

Won't this fail Code Signing checks on the binary? Also, won't not restarting after patching cause problems with unified buffer cache (probably that binary is already loaded and cached)?

@pvaibhav: Actually, the DataDetectorsCore bundle fails Code Signing verification out of the box, so applying this patch will have no additional adverse effects on the validity of the bundle (the error message will change but ultimately it was invalid in the first place). I've been running with this binary patch for at least 12 hours and I've noticed no problems. I'd back up the original version anyway, though, there's no telling what Software Update might do when 10.8.3 comes out.

@landonf: patch for the 32-bit version: printf '\xeb' | dd bs=1 seek=58303 count=1 conv=notrunc of=/System/Library/PrivateFrameworks/DataDetectorsCore.framework/Versions/Current/DataDetectorsCore