Last night, an ALWGW [3] user had another Internet SMTP message arrive that caused a TNOS [4] 2.40 SIGSEGV crash.  The symptoms were the same as those encountered previously (I would recommend reading the background in "Long-term mystery SOLVED [5]," if you haven't already), except that this time there was no X-Face: line present in the header.  The crash did occur in reject.c, but this time it was the To: field causing the problem.  Again, examining the message, no unusual characters were present, although the line was a bit long.

I reviewed and compared the earlier problem messages with the new one, and finally determined what I believe is the key to this issue.  It is not so much the actual content or type of characters that are present in the header lines causing the crashes, but rather the number of characters in any given header line.  It appears to cause a problem when any one line in the headers has more than 256 characters.  When viewed from a 2007 frame of reference, this is a significant weakness!  I will go back and examine the code as I have time, and see if I can come up with a fix.

For now, I have added a line to my Postfix [6] header checks that silently drops any header line exceeding 255 characters:

Many mailers wrap long header lines anyway.  RFC-822 [7] certainly provides for that option.