Re: [MACTCL] more on Tk Cocoa updates

[Kevin W. <kw...@co...>]

On 7/20/12 3:33 PM, Adrian Robert wrote:
> Unfortunately, it seems like everyone who ventures into this code, whatever their skills and determination, ends up heading back out vanquished.  And a complete rewrite is too big of a task to contemplate.

That includes me. It's been awhile since I dove into the code, but as I 
recall, it did involve integration of the Tk and Cocoa event loops at a 
very low level. Daniel Steffen's comment in the thread / bug report that 
Adrian referenced is probably the most astute diagnosis:

"it seems repeated Tcl_ServiceAll() isn't providing quite the semantics
needed here, the idle queue should really be processed only once per
runloop observer invocation (but all other pending tcl events should be
drained)
ISTR trying to achieve this via Tcl_DoOneEvent() invocations initially and
being unsuccessful (git history may provide details), we may need extra
support in tclNotify.c for this."

Daniel is the author of the Cocoa port of Tk, and probably the only 
person qualified to do attack this issue, but as he is not much involved 
in Tk maintenance anymore, I'm afraid we're stuck wtih these issues.

--Kevin

-- 
Kevin Walzer
Code by Kevin
http://www.codebykevin.com