[Dev] vt102 issue?

Zack Salinas zsalinas at alterpoint.com
Sat Apr 5 08:13:46 CDT 2008


I just got this back from someone that was running into some ProCurve backup issues in the commercial software so they did some debugging with adapterTool and came up with this:

*********************************************************
There seem to be a bug ZipTie::SSH, in regard to the handling of vt102.
Every time ZipTie::SSH does a match, it reset its own buffer, but doesn't
reset the vt102 buffer (Term::VT102).  Which gets quite big at the end of
a run, resulting in some nasty stuff when trying to do regex match for
exit strings, especially when the exit subroutine is looking for is part
of the config stuff dumping out by the device (yay, I got a 100Mb log file
from this).  It would also mean that anytime there's an attempt for
command result match, it's matching against ALL command results since,
instead of just issued command (which is the intention from looking at the
code).

I fixed this in the adapter tools, but I'm not sure how to patch it to the
server, would it be the same as uploading plugins?

The fix is to reset the Term::VT102 object every time there is a
buffer_empty() call on the ssh object.

FYI, it looks like this only affect the Procurve with Curses, as
Term::VT102 is use for curses conversion to nicer text (no control chars).
*********************************************************

Has anyone else seen that behavior before?

-Z



More information about the Dev mailing list