Re: clamscan vs. clamscan with mb2md
In a previous msg, I wrote:
Running clamscan over a PDC/BDC with roaming profiles will (obviously) generate sporadic alerts on mbox files assoicated with assorted mail clients,
icedove/tbird in this case. In order to track down the specific message, I've used mbox2maildir (in the past) and mb2md presently to convert them into
a "broken out" situation, i.e. a structure where each message is its own file. I now have a case where the clamscan on the Inbox gives a positive and
clamscan on the mb2md (or mbox2maildir) directory of messages gives a negative. Is this case known? I believe it has occurred for me in the past
(forgotten exactly how long ago) and so it seems to be a neglected bug. However, I'm not sure which package (or support package) is responsible here.
Is clamscan giving a false positive/false negative or is mb2md changing the message in question so that clamscan misses it? It is a user's mailbox and
therefore not properly public for debugging purposes.
The clamscan alert is ".../Inbox: Email.Phishing.Webmail-37 FOUND".
I found some time to track down the offending message in the Inbox and
the only difference wrt causing a clamscan alert or not is the initial
From line on the message. The Inbox had the line and the broken out
mb2md files did not. If I put just that line back into the broken out
message, then the alert returned when scanning the maildir messages.
(This is on a lenny system with clamav 0.96.1+dfsg-1~volatile1, so if it
is a known bug fixed in squeeze, then let me know. thx) I'll go ahead,
if no one objects, and file a bug on clamav since mbox2maildir preserves
a modified form of the from line (prefixes the line with "MBOX-Line: ")
but still doesn't trigger a clamscan alert.