[Tcl-bugs] [ tcl-Bugs-2093947 ] coroutine+unset trace misfeature

[SourceForge.net <no...@so...>]

Bugs item #2093947, was opened at 2008-09-04 19:48
Message generated for change (Comment added) made by msofer
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=110894&aid=2093947&group_id=10894

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: None
Group: development: 8.6a3
Status: Open
>Resolution: Fixed
Priority: 5
Private: No
Submitted By: Donal K. Fellows (dkf)
>Assigned to: Donal K. Fellows (dkf)
Summary: coroutine+unset trace misfeature

Initial Comment:
I've been experimenting with traces and coroutines as part of the discussion on http://wiki.tcl.tk/21537 (as to why Lars's proposal isn't good enough) and I came across this:

$ make shell
% proc foo {} {set v 1; trace add variable v {write unset} bar; yield; set v 2; yield; set v 3}
% proc bar args {puts $args};namespace path tcl::unsupported;coroutine a foo
% a
v {} write
% a
v {} write
v {} unset
3
% a
ambiguous command name "a": after append apply array atProcExit auto_execok auto_import auto_load auto_load_index auto_qualify
% coroutine a foo
% a
v {} write
% rename a {}
% a
ambiguous command name "a": after append apply array atProcExit auto_execok auto_import auto_load auto_load_index auto_qualify

It would be *very* nice if variable unset traces were run when a coroutine is "deleted early" since that would mean that the assumption that unset traces could be used to handle scoping of resource lifetime to procedure body execution lifetime was guaranteed. And in a strong sense, the variables are unset when the created coroutine is deleted.

There's probably also a related issue with tailcall not calling unset traces on local variables, but I've not checked. ;-)

I'm sorry these things make stuff messier. Resolving them would, I think, go a long way towards making the NRE stuff fit to come out of the tcl::unsupported closet.

----------------------------------------------------------------------

>Comment By: miguel sofer (msofer)
Date: 2008-09-04 22:22

Message:
Logged In: YES 
user_id=148712
Originator: NO

Fixed in HEAD: insure that unset traces ignore the "execEnvPtr->rewind"
that signals coroutine deletion.

Leaving open and reassigning: should the same thing be done for unwinding
when interpreter limits are violated? When unwinding the interp? Possibly
... possibly not.

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=110894&aid=2093947&group_id=10894