The procedure until is not in the core, but you can define it yourself as follows: ====== proc until {cond code} { uplevel 1 [list while !($cond) $code] } ====== See also [control]::do (in [tcllib]). [PL]: The original rewriting of the command was like this: === list while !$cond $code === but that doesn't work since a `cond` value of, say, `$i > 5` is rewritten as the condition argument `!$i > 5` which will always evaluate to false (the `!` operator has priority, and the result of evaluating it is either 1 or 0, which is never greater than 5). Other condition strings are likely to lead to other interesting errors. Rewriting the command as === list while !($cond) $code === ensures that all of `$cond` is evaluated before negating the result. ---- This doesn't implement the usual meaning of ''until'', which executes the code and then tests cond after. [Larry Smith] I use: # an unbounded loop or one with a terminating condition of while or until. # always executes at least once. If no while or until conditions is attached, # loops forever until . proc repeat {args} { > script whenexit test if {[catch {^^ $script} err]} ^ case $whenexit { "" {^^ "while 1 \{$script\}"} "while" {^^ "while \{$test\} \{$script\}"} "until" {^^ "while \{!($test)\} \{$script\}"} } } The advantage of the above is you can leave out the while or until entirely and just use break to exit from anywhere in the loop. [DKF]: The usual definition is typically expressed as a do-while loop, which is a form that puts the test last. When expressed condition-first, it is natural to expect the condition to be checked first... [RLH] Someone else was thinking about this too: http://christopherchase.cc/posts/tcl-until.html%|%until%|% <> Control Structure