Tcl Source Code

Name resolution for variablesin namespaces sometimes
finds variables that "do not exist" according to [info
vars].

  % namespace eval a upvar x q
  % info vars x
  % namespace eval b set x 1
  1
  % set x
  1
 
  % trace add variable x write {;#}
  % info vars x
  % namespace eval a set x 1
  1
  % set x
  1

  % namespace eval a {trace add variable x write {;#}}
  % set x 0
  0
  % info vars a::x
  % namespace eval a set x 1
  1
  % set x
  0
  % info vars a::x
  ::a::x

The variables which are found contrary to [info vars]
satisfy:
 - they have an entry in their namespace's hash table
 - they are unset, ie (varPtr->flags & VAR_UNDEFINED)
 - they are not namespace variables, ie 
       !(varPtr->flags & VAR_NAMESPACE_VAR)

Logged In: YES 
user_id=148712

Another surprise:
% namespace eval a set x
can't read "x": no such variable
% set x 1
1
% namespace eval a set x
1
% upvar 0 ::a::x y
% namespace eval a set x
can't read "x": no such variable
%

Logged In: YES 
user_id=148712

As many (?) scripts and apps (at least, AlphaTk: see bug
#959786) rely on the details of the definition of
"existence" for variable name resolution, they break with
any change in these details.

Any action on this bug will have to wait. The patch was
reverted from HEAD, and should not be applied in the 8.4
line either.

Logged In: YES 
user_id=148712

Note that the final judge of which variable should be set is 
[namespace which -variable], according to namespace.n 
This is not changed by the patch. 

The patch insures that 
   set name [namespace which -variable x]
   info vars $name
always returns $name. Previously:

  % trace add variable x write {;#}
  % namespace eval a {
          set name [namespace which -variable x]
      }
  ::x
  %  info vars ::x

File Added - 88161: 959052.patch

Logged In: YES 
user_id=148712

Attaching a patch, already committed to HEAD. Please review
for itcl compat. Are there any other known users of
Tcl_FindNamespaceVar that may rely on the weird lookup
semantics?

Second Q: should this be backported to 8.4? 8.3?