| Author |
Tcl/Tk 8.5.0 Release Candidates
|
|
| Donald G Porter 2007-12-10, 7:17 pm |
|
We are in the closing days of development leading to the release of
Tcl/Tk 8.5.0. "Release Candidate 2" [*] is available via ftp:
ftp://ftp.tcl.tk/pub/tcl/tcl8.5.0rc2-src.tar.gz
ftp://ftp.tcl.tk/pub/tcl/tk8.5.0rc2-src.tar.gz
ftp://ftp.tcl.tk/pub/tcl/tcl8.5.0rc2-html.tar.gz
for testing purposes. Please report any issues to the SourceForge
Trackers without delay. Please limit the use of these to testing.
Do not embed them in products for distribution to end users. Save
that for the actual 8.5.0 releases.
There will be at least one more Release Candidate that may become
the 8.5.0 release. At a minimum, some changes files and release note
matters need more updates.
[*] Release Candidates 0 and 1 were announced on developer mailing lists
only.
--
| Don Porter Mathematical and Computational Sciences Division |
| donald.porter@nist.gov Information Technology Laboratory |
| http://math.nist.gov/~DPorter/ NIST |
|_______________________________________
_______________________________|
| |
|
|
|
|
|
|
| Greg Couch 2007-12-13, 10:14 pm |
| "Larry W. Virden" <lvirden@gmail.com> writes:
>I do hope that anyone who is planning on upgrading to Tcl/Tk 8.5 (with
>the new gui widget set, etc.) has checked out at least one of the
>release candidate releases and reported problems encountered.
I tried 8.5rc2 in our application that uses Tk via Python, and I'm seeing
some weird behavior, like text widgets where the text isn't selectable and
sometimes you can't type into it, but if I iconify and deiconify the dialog,
then I can type in it again (but the text is still unselectable).
I also noticed that for the Tk Aqua version, that 8.4.16 and 8.5 changed
something with threads and our Python application regularly crashes with
a treading error, so we've dropped the Aqua version.
In both cases, I haven't had time yet to dig in and figure what's really
wrong. But I will be monitoring the newsgroup and reading about other
folks' experiences with 8.5.
Greg Couch
UCSF Computer Graphics Lab
| |
| George Peter Staplin 2007-12-13, 10:14 pm |
| Greg Couch wrote:
> "Larry W. Virden" <lvirden@gmail.com> writes:
>
>
> I tried 8.5rc2 in our application that uses Tk via Python, and I'm seeing
> some weird behavior, like text widgets where the text isn't selectable and
> sometimes you can't type into it, but if I iconify and deiconify the dialog,
> then I can type in it again (but the text is still unselectable).
Can you show a simple test case? I wonder what is causing this. I know
that <<Modified>> is being fixed in the Tk text widget today, and
yesterday. There's also the bug I posted about with Tk_RedrawImage,
which looks like it may not be fixed by the time of 8.5.0.
> I also noticed that for the Tk Aqua version, that 8.4.16 and 8.5 changed
> something with threads and our Python application regularly crashes with
> a treading error, so we've dropped the Aqua version.
Are you using Tk in more than 1 thread? It doesn't work well in
general to use Tk in more than thread. A test case for this would also
be useful. What is the threading error you see?
> In both cases, I haven't had time yet to dig in and figure what's really
> wrong. But I will be monitoring the newsgroup and reading about other
> folks' experiences with 8.5.
I once generated some statistics years ago, and found that a large
portion of the Tcl and Tk sources have changed between the 8.4 and 8.5
release > 50% IIRC. It's probably even larger now. My statistics were
meant to push the Tcl Core Team for more releases, but Jeff Hobbs at the
time felt there weren't enough new features to warrant a release.
I'm personally hoping that 8.5/8.6 is eventually a better release than
8.4, but most people don't run the HEAD, and with the core team being
uhm difficult it's frustrating.
Note: I'm not condemning all of the core team. DGP has certainly been
pushing for releases, and replying to 4 year old bug reports assigned to
TCT members that seemingly don't have the time. KBK has wanted 9.0 for
years.
In my opinion, the Tcl/Tk core test suites give *some* TCT developers a
false sense of security. It's very difficult to fully test code paths,
especially when so many changes have occured since 8.4.
You may recall that I was searching for more Tcl developers to fix bugs
years ago, and posting on comp.lang.tcl to motivate people. Nothing
came of it, and I was highly disappointed. So some of you have no one
to blame but yourselves. Shame on you :)
George
| |
| George Peter Staplin 2007-12-14, 10:16 pm |
| George Peter Staplin wrote:
> Greg Couch wrote:
>
> I was kind of hoping to nudge you for more info, and I got it. I'll try
> to build a simple test case using Python and the latest Tk.
>
> I'm familiar with IDLE. It's the editor extraordinaire for Python that
> comes with Python and uses Tkinter. IIRC Guido van Rossum wrote most of
> it.
>
> I can't be of help with the MacOS X problems yet.
I can duplicate the bug with 8.5/HEAD of Tcl and Tk using the latest
Python from python.org with IDLE. I'm trying to track it down now.
George
| |
|
|
| Daniel A. Steffen 2007-12-15, 8:12 am |
| On Dec 15, 12:06 am, Greg Couch <gr...@cgl.ucsf.edu> wrote:
> As for Aqua, we only use Tk in one thread because Python enforces that
> automatically. I don't have access to Mac OS X today, but the error was
> something like: "Fatal Python error: PyEval_ReleaseThread: NULL thread state"
> and then Python exited. That indicates something changed with threading
> on Aqua since the identical code is used for Windows and X11 and it works
> on those platforms (and for Aqua in Tk 8.4.15).
just checked, I see nothing thread related specific to Aqua that has
changed since 8.4.15, so your report is puzzling; the
PyEval_ReleaseThread message could of course be a red herring and the
problem may be caused by other changes.
Are you running against a thread-enabled build of Tk (& Tcl)? Unlike
on other unix platforms, the threaded build of Tcl/Tk is usually the
default for binaries on Mac OS X (for historical reasons), that may be
a point of difference between X11 (on other unix platforms) and
Aqua... If yes, you could try running against a TkAqua configured with
--disable-threads.
If you have a small testcase that reproduces this problem, please post
a Tk bug report at sourceforge and I'll take a look ASAP, c.l.t is not
a very reliable avenue for reporting bugs...
Cheers,
Daniel
| |
| Uwe Klein 2007-12-15, 8:12 am |
| George Peter Staplin wrote:
[color=darkred]
I found this to be an issue with mozilla and friends as well.
( You have to open another window to get your focus back.
Keyboard input to another window than your mouse events.
)
IMHO it is connected with the XIM module of the X-Server.
uwe
| |
|
| Dnia 15.12.2007 Don Porter <dgporter@verizon.net> napisa³/a:
> Even before some of the problems with Python's Tkinter were
> noted here, some other important post-RC3 commits came in.
And this "rename send {}"-related problem (still here) should be considered
as harmless?
--
ZB
| |
| miguel 2007-12-15, 7:15 pm |
| ZB wrote:
> Dnia 15.12.2007 Don Porter <dgporter@verizon.net> napisa³/a:
>
>
> And this "rename send {}"-related problem (still here) should be considered
> as harmless?
Is it this bug ticket
http://sourceforge.net/tracker/inde...97&atid=112997?
The maintainer says: "This is a real bug, and ought to be fixed before
8.5.0, but it's not severe: the worst end-user consequence is a
confusing/annoying error message at program shutdown in (probably-) rare
circumstances." He also said priority 8 (9 would be "release blocker").
| |
|
| Dnia 15.12.2007 miguel <msofer@users.sf.net> napisa³/a:
> confusing/annoying error message at program shutdown in (probably-) rare
> circumstances
Perhaps it could be possible to block that error message? F.e. by directing
it to /dev/null/ or sth.?
--
ZB
| |
| miguel 2007-12-15, 7:15 pm |
| ZB wrote:
> Dnia 15.12.2007 miguel <msofer@users.sf.net> napisa³/a:
>
>
> Perhaps it could be possible to block that error message? F.e. by directing
> it to /dev/null/ or sth.?
Please take the discussion of the bug to its proper venue: the ticket
tracker.
I am myself completely unqualified to comment on your proposal.
| |
|
| Dnia 15.12.2007 miguel <msofer@users.sf.net> napisa³/a:
>
> Please take the discussion of the bug to its proper venue: the ticket
> tracker.
>
> I am myself completely unqualified to comment on your proposal.
No, it's not about fixing. I'm wondering, if I can block that message for
now. Perhaps you know a way?
--
ZB
| |
| Donal K. Fellows 2007-12-15, 7:15 pm |
| Uwe Klein wrote:
> IMHO it is connected with the XIM module of the X-Server.
There are bugs in Tk relating to XIM handling; it's one of the less
well documented part of X (since most people writing code just say
"call this all it work" and don't give much guidance beyond very
cryptic source...) so working out what assumptions are safe and how to
merge in other event processing loops is tricky.
Donal.
| |
|
| Dnia 15.12.2007 Don Porter <dgporter@verizon.net> napisa³/a:
> Even before some of the problems with Python's Tkinter were
> noted here, some other important post-RC3 commits came in.
> [..]
> The real 8.5.0 releases are still to come. Best guess right
> now is early-to-mid-next w .
Wait, I've found a real one (using RC2): the dimensions of the radiobuttons
and checkbuttons (both "normal" and themed ones) widgets are no longer
adjusted according to the size of the chosen font. Checked it many times.
Perhaps some other widgets needs such check-out.
--
ZB
| |
| Cesar Rabak 2007-12-15, 7:15 pm |
| ZB escreveu:
> Dnia 15.12.2007 Don Porter <dgporter@verizon.net> napisa³/a:
>
>
> Wait, I've found a real one (using RC2): the dimensions of the radiobuttons
> and checkbuttons (both "normal" and themed ones) widgets are no longer
> adjusted according to the size of the chosen font. Checked it many times.
>
> Perhaps some other widgets needs such check-out.
I think this brings an interesting discussion:
Should de widget adjust to the font (size) or should it be around?
Better, as it may show we have two schools of thought on this, shouldn't
we have a flag to choose either of these behaviours?
just my .0199999...
Regards,
--
Cesar Rabak
| |
|
| Dnia 15.12.2007 Cesar Rabak <csrabak@yahoo.com.br> napisa³/a:
> I think this brings an interesting discussion:
>
> Should de widget adjust to the font (size) or should it be around?
Discussion?
Just make a simple test:
pack [radiobutton .r -text "Hi" -font "Arial 20"]
Do you like it?
> Better, as it may show we have two schools of thought on this, shouldn't
> we have a flag to choose either of these behaviours?
Perhaps additional option.
--
ZB
| |
|
| Dnia 15.12.2007 Cesar Rabak <csrabak@yahoo.com.br> napisa³/a:
> Should de widget adjust to the font (size) or should it be around?
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
OK, just few words, why: just try to run your own programs using the screen
of much different (perhaps very high) resolution. The fonts size can be
adjusted with not so much effort - but how will all that check- und
radiobuttons look like, when the resizeability has been blocked?
--
ZB
| |
|
|
| Kevin Walzer 2007-12-17, 10:21 pm |
| George Peter Staplin wrote:
> George Peter Staplin wrote:
>
> I can duplicate the bug with 8.5/HEAD of Tcl and Tk using the latest
> Python from python.org with IDLE. I'm trying to track it down now.
>
>
I can reproduce these issues with IDLE and Tk 8.5 RC 5 on Mac OS X
10.5.1. Here is the strange behavior that I observe:
1. Toplevel windows do not take focus unless selected with the menu.
2. Text is not selectable.
3. Cannot set insertion point with mouse.
--
Kevin Walzer
Code by Kevin
http://www.codebykevin.com
| |
| George Peter Staplin 2007-12-17, 10:21 pm |
| Kevin Walzer wrote:
> George Peter Staplin wrote:
>
> I can reproduce these issues with IDLE and Tk 8.5 RC 5 on Mac OS X
> 10.5.1. Here is the strange behavior that I observe:
>
> 1. Toplevel windows do not take focus unless selected with the menu.
> 2. Text is not selectable.
> 3. Cannot set insertion point with mouse.
Tkinter has had some problems for a while. During my tests today I
found that it fails going all the way back to 8.5a1. I'm still not sure
what the cause is, but I've noted most of the information in this bug
report:
http://sourceforge.net/tracker/inde...997&atid=112997
I also reported the bug to python-dev@python.org, and someone said he'd
try to look into it this w. We will see what happens.
I'd appreciate any help from the community in tracking it down further,
because Tkinter support is important to us. The bug is also assigned to
me, so I'm hoping it won't get lost in the heap.
George
| |
| Kevin Walzer 2007-12-18, 7:20 pm |
| George Peter Staplin wrote:
> Kevin Walzer wrote:
>
> Tkinter has had some problems for a while. During my tests today I
> found that it fails going all the way back to 8.5a1. I'm still not sure
> what the cause is, but I've noted most of the information in this bug
> report:
>
> http://sourceforge.net/tracker/inde...997&atid=112997
>
>
> I also reported the bug to python-dev@python.org, and someone said he'd
> try to look into it this w. We will see what happens.
>
> I'd appreciate any help from the community in tracking it down further,
> because Tkinter support is important to us. The bug is also assigned to
> me, so I'm hoping it won't get lost in the heap.
>
Thanks for looking into this. I'm glad you reported the bug to the
Python developers as well--I checked the py-dev archives and Martin V.
Lowis is the right person to ask. From what I've observed, it does look
like the issues stem from interaction between the Python and Tk
libraries, so there probably are some fixes that need to be implemented
on the Python side. (I haven't seen bugs like this in Tcl/Tk itself,
i.e. the Wish 8.5 widget demo.) I develop in Python/Tkinter as well and
use IDLE extensively, so these problems are frustrating.
--
Kevin Walzer
Code by Kevin
http://www.codebykevin.com
| |
|
| George Peter Staplin wrote:
>
> I was kind of hoping to nudge you for more info, and I got it. I'll try
> to build a simple test case using Python and the latest Tk.
>
> I'm familiar with IDLE. It's the editor extraordinaire for Python that
> comes with Python and uses Tkinter. IIRC Guido van Rossum wrote most of
> it.
>
> I can't be of help with the MacOS X problems yet.
FWIW, maybe this helps in the discussion: we found that in our OpenMP
Aqua application which uses an integrated Tcl 8.4.16 interpreter we
needed to disable the threads in order to avoid hangs. We didn't try 8.5
yet..
-- Hans
| |
| Tobias Hippler 2007-12-19, 4:28 am |
| Hi,
I did a make test run on Windows XP for tcl8.5.0rc5.
One test failed:
socket.test
==== socket-2.11 detecting new data FAILED
==== Contents of test case:
proc accept {s a p} {
global sock
set sock $s
}
set s [socket -server accept 0]
set sock ""
set s2 [socket 127.0.0.1 [lindex [fconfigure $s -sockname] 2]]
vwait sock
puts $s2 one
flush $s2
after 500
fconfigure $sock -blocking 0
set result a:[gets $sock]
lappend result b:[gets $sock]
fconfigure $sock -blocking 1
puts $s2 two
flush $s2
fconfigure $sock -blocking 0
lappend result c:[gets $sock]
fconfigure $sock -blocking 1
close $s2
close $s
close $sock
set result
---- Result was:
a:one b: c:
---- Result should have been (exact matching):
a:one b: c:two
==== socket-2.11 FAILED
| |
|
| It amuses me that this thread-that-will-not-die is juxtaposed with a
series in Dilbert (part of my morning read) in which the pointy haired
boss has hired a dead horse to beat.
Larry Wall is Larry Wall. Love him or not (and I love him for what he
is and what he's done), we can't expect him to be anyone else but
Larry Wall. Forgive and move on? Not a bad plan. Respond on his list,
politely bringing them up to date? I could see that. Complain among
ourselves? Why?
I love Tcl/Tk and this community for its brilliance, its generosity
(to me at least), and its ability to put language issues before
personalities. But not everyone knows about that.
If we want to discuss outreach (yet again), I'm in. More Larry Wall
may be an exercise in felling trees in an empty forest.
~Lan
| |
| Bryan Oakley 2007-12-19, 7:16 pm |
| Lan wrote:
> It amuses me that this thread-that-will-not-die is juxtaposed with a
> series in Dilbert (part of my morning read) in which the pointy haired
> boss has hired a dead horse to beat.
>
> Larry Wall is Larry Wall. Love him or not (and I love him for what he
> is and what he's done), we can't expect him to be anyone else but
> Larry Wall. Forgive and move on? Not a bad plan. Respond on his list,
> politely bringing them up to date? I could see that. Complain among
> ourselves? Why?
That therad looooonnnnng ago stopped talking about Larry Wall and nobody
thought to change the subject.
--
Bryan Oakley
http://www.tclscripting.com
| |
| Donald G Porter 2007-12-19, 7:16 pm |
| Donald G Porter wrote:
> Once again, intent is to rename these files to the real 8.5.0 release
> tomorrow, unless some other serious bug fix comes in.
Serious bug fixes have in fact come in. Now we are at Release
Candidate 7 for Tcl 8.5.0.
ftp://ftp.tcl.tk/pub/tcl/tcl8_5/tcl8.5.0rc7-src.tar.gz
ftp://ftp.tcl.tk/pub/tcl/tcl8_5/tcl850rc7-src.zip
The other 8.5.0 RC files remain unchanged at RC 5.
--
| Don Porter Mathematical and Computational Sciences Division |
| donald.porter@nist.gov Information Technology Laboratory |
| http://math.nist.gov/~DPorter/ NIST |
|_______________________________________
_______________________________|
| |
|
|
| Donal K. Fellows 2007-12-20, 10:13 pm |
| Andrés García wrote:
> So, are the 8.5 files at SourceForge the final release?
Wait for the announcement, Mr. Impatient! ;-)
Donal.
|
|
|
|