Tcl Library Source Code

View Ticket
Login
Bounty program for improvements to Tcl and certain Tcl packages.
Ticket UUID: 3574004
Title: decrypt/encrypt fails for data beginning with '-'
Type: Bug Version: None
Submitter: samoconnor Created on: 2012-10-03 02:28:14
Subsystem: aes Assigned To: andreas_kupries
Priority: 5 Medium Severity:
Status: Closed Last Modified: 2013-01-09 05:38:04
Resolution: Fixed Closed By: andreas_kupries
    Closed on: 2013-01-08 22:37:21
Description:
% set data_block -[string repeat \\0 15]
-\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0
% aes::aes -hex -mode cbc -dir encrypt -key $nil_block $data_block
bad option "-\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0": must be one of -chunksize, -dir, -hex, -in, -iv, -key, -mode, -out

The last item in the args list is always the data (or the in stream name).
It is never a "-" option and should not be treated as one even if it starts with a "-".

I have a server system that has been running for more than a year before happening to have some data beginning with a "-".
Very hard to test for unless you already know what the problem is.

Suggested fix: terminate argument processing when the args list length gets to 1...

--- tcllib-1.14.orig/modules/aes/aes.tcl2012-07-25 10:53:21.000000000 +1000
+++ tcllib-1.14/modules/aes/aes.tcl2012-10-03 12:19:44.000000000 +1000
@@ -519,7 +519,7 @@
     set opts(-iv) [string repeat \0 16]
     set modes {ecb cbc}
     set dirs {encrypt decrypt}
-    while {[string match -* [set option [lindex $args 0]]]} {
+    while {[llength $args] > 1 && [string match -* [set option [lindex $args 0]]]} {
         switch -exact -- $option {
             -mode      { set opts(-mode) [SetOneOf $modes [Pop args 1]] }
             -dir       { set opts(-dir) [SetOneOf $dirs [Pop args 1]] }
User Comments: andreas_kupries added on 2013-01-09 05:38:04:
Committed revision is [57b5f729ca].

andreas_kupries added on 2013-01-09 05:37:21:
allow_comments - 1

Decided to accept the change, as it should be pretty much backward compatible.
Additionally documented the option -- also.
Version bumped to 1.1. (Visible API change).
I am lucky that this package does not have C accelerator code requiring the change also.
That said, getting some C code here for speed would be nice.

andreas_kupries added on 2013-01-09 05:22:40:
The aes::aes command already accepts a (currently undocumented) option -- (double-dash) which stops option processing and forces interpretation of data, even if starting with a dash. See line 532, about 10 lines after where you patched.

Is that good enough for your case (plus documenting this option) ?