[Tcl-bugs] [ tcl-Bugs-217934 ] Tcl_AsyncMark doesn't appear to wake the notifier under UNIX

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

Bugs item #217934, was opened at 2000-10-25 17:46
You can respond by visiting: 
http://sourceforge.net/tracker/?func=detail&atid=110894&aid=217934&group_id=10894

Category: 04. Async Events
Group: = 8.2.1
Status: Closed
Resolution: Fixed
Priority: 2
Submitted By: Nobody/Anonymous (nobody)
Assigned to: Nobody/Anonymous (nobody)
Summary: Tcl_AsyncMark doesn't appear to wake the notifier under UNIX

Initial Comment:
OriginalBugID: 3423 RFE
Version: 8.2.1
SubmitDate: '1999-11-05'
LastModified: '2000-04-04'
Severity: MED
Status: Closed
Submitter: techsupp
ChangedBy: hobbs
RelatedBugIDs: 3862 4826
OS: Windows NT
OSVersion: 5.00.2128
FixedDate: '2000-10-25'
ClosedDate: '2000-04-04'


Name:
David Gravereaux

Comments:
Could someone verify this?

DesiredBehavior:
TclpAsyncMark is #define'd out in tclUnixPort.h and thus has no way to awake select() in Tcl_WaitForEvents (should the notifier be blocked).  I can't verify that this is true as I only spotted it reading the source and don't program for unix.

If this is true, I was wondering if a TclpAsyncMark could be added that would generate a proper signal to awake the blocking select?





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

>Comment By: David Gravereaux (davygrvy)
Date: 2001-12-04 14:38

Message:
Logged In: YES 
user_id=7549

This is fixed in 8.4a2+ (around august 2000) and was also 
merged down to 8.3.4 (around august 2001).


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

Comment By: Dave Thomas (davethomas)
Date: 2001-12-04 07:18

Message:
Logged In: YES 
user_id=393111



We have been using a work-around for this bug for a couple 
of years now.  On unix, the notifier is not awakened by a 
Tcl_AsyncMark if it's blocked in a select statement.  For 
example, if the application is driven by socket events, 
events posted using Tcl_AsyncMark won't be executed until 
the next socket readable event occurs.

So on unix, we always do a dummy write to a socket that 
we've bound a "dummy" handler.  The handler simply reads 
the socket and returns, but then the notifier processes 
the "real" event that we posted via Tcl_AsyncMark.

But this work around DOUBLES, the number of socket events 
that would otherwise need to be handled.

We're still using 8.3.0.  I can find no indication that 
this bug has been fixed in later releases.  What is the 
status?  Is there a fix or patch?

Thanks,

Dave Thomas

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

Comment By: Nobody/Anonymous (nobody)
Date: 2001-06-29 11:56

Message:
Logged In: NO 

Has this bug been fixed?  If so, what release.

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

You can respond by visiting: 
http://sourceforge.net/tracker/?func=detail&atid=110894&aid=217934&group_id=10894