From: Ned D. <na...@ac...> - 2013-10-23 20:06:36
|
Now that OS 10.9 Mavericks has been released (and at no cost), people are updating to it and running into a problem with Aqua Tk. I ran into the problem with Python's IDLE but it seems to affect other multi-window Tk applications, as it can be reproduced using wish demos. Basically, the problem is that Tk windows don't get automatically redrawn when the mouse is used to change the active window and focus. I've described the details here: https://core.tcl.tk/tk/tktview?name=53f7a1b553 As it affects current versions of ActiveTcl 8.5 and 8.6, I also opened an ActiveTcl issue: http://bugs.activestate.com/show_bug.cgi?id=101210 As noted in the Tk issue above, the Apple-supplied Tk 8.5.9 does not have this problem but that version has other, serious problems that have been fixed in newer versions of Tk 8.5, for example, immediately crashing when typing a composing character in a text field (like option-u for US Extended input methods). Also as noted, one workaround appears to be to force Tcl/Tk to run in 32-bit mode, which may not be possible for some applications if they depend on non-universal libraries. -- Ned Deily, na...@ac... |
From: Kevin W. <kw...@co...> - 2013-10-24 21:54:49
|
On 10/23/13 4:06 PM, Ned Deily wrote: > Now that OS 10.9 Mavericks has been released (and at no cost), people are > updating to it and running into a problem with Aqua Tk. I ran into the > problem with Python's IDLE but it seems to affect other multi-window Tk > applications, as it can be reproduced using wish demos. Basically, the > problem is that Tk windows don't get automatically redrawn when the mouse is > used to change the active window and focus. I hope to install Mavericks this weekend and I'll take a look then. --Kevin -- Kevin Walzer Code by Kevin/Mobile Code by Kevin http://www.codebykevin.com http://www.wtmobilesoftware.com |
From: Adrian R. <adr...@gm...> - 2013-10-25 04:23:33
|
I’ve just installed it and Tk-Cocoa is a mess under Mavericks. Something with the new timer coalescing interacts very badly with the incorrectly-implemented event and drawing loops in Tk-Cocoa. A quick fix might be to use the new timer APIs to request low jitter. But one might hope that the bugs themselves might help point out more clearly what is wrong with the current impl. -Adrian On Oct 25, 2013, at 12:54 AM, Kevin Walzer <kw...@co...> wrote: > On 10/23/13 4:06 PM, Ned Deily wrote: >> Now that OS 10.9 Mavericks has been released (and at no cost), people are >> updating to it and running into a problem with Aqua Tk. I ran into the >> problem with Python's IDLE but it seems to affect other multi-window Tk >> applications, as it can be reproduced using wish demos. Basically, the >> problem is that Tk windows don't get automatically redrawn when the mouse is >> used to change the active window and focus. > > I hope to install Mavericks this weekend and I'll take a look then. > > --Kevin > > > > -- > Kevin Walzer > Code by Kevin/Mobile Code by Kevin > http://www.codebykevin.com > http://www.wtmobilesoftware.com > > ------------------------------------------------------------------------------ > October Webinars: Code for Performance > Free Intel webinars can help you accelerate application performance. > Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from > the latest Intel processors and coprocessors. See abstracts and register > > http://pubads.g.doubleclick.net/gampad/clk?id=60135991&iu=/4140/ostg.clktrk > _______________________________________________ > Tcl-mac mailing list > tc...@li... > https://lists.sourceforge.net/lists/listinfo/tcl-mac |
From: Kevin W. <kw...@co...> - 2013-10-25 11:07:17
|
On 10/25/13 12:23 AM, Adrian Robert wrote: > I’ve just installed it and Tk-Cocoa is a mess under Mavericks. Something with the new timer coalescing interacts very badly with the incorrectly-implemented event and drawing loops in Tk-Cocoa. A quick fix might be to use the new timer APIs to request low jitter. But one might hope that the bugs themselves might help point out more clearly what is wrong with the current impl. > > -Adrian I was reading about those new API's in Siracusa's review and wondered what their effect might be. Sounds like a mess indeed. As noted, I'll try to install Mavericks this weekend (I have four machines plus a server to upgrade), and will do some more digging after that. --Kevin -- Kevin Walzer Code by Kevin/Mobile Code by Kevin http://www.codebykevin.com http://www.wtmobilesoftware.com |
From: Steve A <ste...@gm...> - 2013-10-25 12:32:44
|
Our app checks for "-psn" in the command line to detect if called as an App, but this no longer seems to be emplyed ?? > I was reading about those new API's in Siracusa's review and wondered > what their effect might be. Sounds like a mess indeed. Hmmm - just what we need for our timer based problems. Timer coalescing is probably going to take a lot of debugging for performance critical apps. FWIW, our 32bit carbon 8.5.9 framework seems to go ok on the last mavericks beta (13a598).... OpenGL/game performance seem to be affected by the OS upgrade too. http://forums.macrumors.com/showthread.php?p=18196922#post18196922 On Fri, Oct 25, 2013 at 9:07 PM, Kevin Walzer <kw...@co...> wrote: > On 10/25/13 12:23 AM, Adrian Robert wrote: >> I’ve just installed it and Tk-Cocoa is a mess under Mavericks. Something with the new timer coalescing interacts very badly with the incorrectly-implemented event and drawing loops in Tk-Cocoa. A quick fix might be to use the new timer APIs to request low jitter. But one might hope that the bugs themselves might help point out more clearly what is wrong with the current impl. >> >> -Adrian > > I was reading about those new API's in Siracusa's review and wondered > what their effect might be. Sounds like a mess indeed. > > As noted, I'll try to install Mavericks this weekend (I have four > machines plus a server to upgrade), and will do some more digging after > that. > > --Kevin > > -- > Kevin Walzer > Code by Kevin/Mobile Code by Kevin > http://www.codebykevin.com > http://www.wtmobilesoftware.com > > ------------------------------------------------------------------------------ > October Webinars: Code for Performance > Free Intel webinars can help you accelerate application performance. > Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from > the latest Intel processors and coprocessors. See abstracts and register > > http://pubads.g.doubleclick.net/gampad/clk?id=60135991&iu=/4140/ostg.clktrk > _______________________________________________ > Tcl-mac mailing list > tc...@li... > https://lists.sourceforge.net/lists/listinfo/tcl-mac |
From: Kevin W. <kw...@co...> - 2013-10-27 20:46:54
|
Hi Ned, On 10/23/13, 4:06 PM, Ned Deily wrote: > Now that OS 10.9 Mavericks has been released (and at no cost), people are > updating to it and running into a problem with Aqua Tk. I ran into the > problem with Python's IDLE but it seems to affect other multi-window Tk > applications, as it can be reproduced using wish demos. Basically, the > problem is that Tk windows don't get automatically redrawn when the mouse is > used to change the active window and focus. I've described the details here: > > https://core.tcl.tk/tk/tktview?name=53f7a1b553 > > As it affects current versions of ActiveTcl 8.5 and 8.6, I also opened an > ActiveTcl issue: > > http://bugs.activestate.com/show_bug.cgi?id=101210 > > As noted in the Tk issue above, the Apple-supplied Tk 8.5.9 does not have this > problem but that version has other, serious problems that have been fixed in > newer versions of Tk 8.5, for example, immediately crashing when typing a > composing character in a text field (like option-u for US Extended input > methods). Also as noted, one workaround appears to be to force Tcl/Tk to run > in 32-bit mode, which may not be possible for some applications if they depend > on non-universal libraries. Daniel Steffen provided me a patch for the issue that he devised during the development of Mavericks, and which he applied to Apple's private branch of Tk-Cocoa (essentially the old 8.5.9 branch, as you noted in the bug report). Daniel's patch is why the behavior isn't visible in Apple's system-provided Tk, because it was fixed already. :-) Now that Mavericks is out he was able to provide the patch for upstream use. After installing Mavericks and testing with my existing build of Tk, I did see the behavior. After doing a fresh checkout of trunk and 8.5, applying the patch, and rebuilding Tcl/Tk, I don't see any of the behavior indicated. As a result I believe this patch solves the problem and I have committed it. Thanks again for the report, and thanks to Daniel for the patch. Best, Kevin -- Kevin Walzer Code by Kevin/Mobile Code by Kevin http://www.codebykevin.com http://www.wtmobilesoftware.com |
From: Jeff H. <je...@ac...> - 2013-10-27 22:00:20
|
We are building updated ActiveTcl installations for 8.5 and 8.6 with this patch this weekend. If someone wants to test from that, let me know. Jeff On 2013-10-27, at 1:46 PM, Kevin Walzer <kw...@co...> wrote: > Hi Ned, > > On 10/23/13, 4:06 PM, Ned Deily wrote: >> Now that OS 10.9 Mavericks has been released (and at no cost), people are >> updating to it and running into a problem with Aqua Tk. I ran into the >> problem with Python's IDLE but it seems to affect other multi-window Tk >> applications, as it can be reproduced using wish demos. Basically, the >> problem is that Tk windows don't get automatically redrawn when the mouse is >> used to change the active window and focus. I've described the details here: >> >> https://core.tcl.tk/tk/tktview?name=53f7a1b553 >> >> As it affects current versions of ActiveTcl 8.5 and 8.6, I also opened an >> ActiveTcl issue: >> >> http://bugs.activestate.com/show_bug.cgi?id=101210 >> >> As noted in the Tk issue above, the Apple-supplied Tk 8.5.9 does not have this >> problem but that version has other, serious problems that have been fixed in >> newer versions of Tk 8.5, for example, immediately crashing when typing a >> composing character in a text field (like option-u for US Extended input >> methods). Also as noted, one workaround appears to be to force Tcl/Tk to run >> in 32-bit mode, which may not be possible for some applications if they depend >> on non-universal libraries. > > Daniel Steffen provided me a patch for the issue that he devised during > the development of Mavericks, and which he applied to Apple's private > branch of Tk-Cocoa (essentially the old 8.5.9 branch, as you noted in > the bug report). Daniel's patch is why the behavior isn't visible in > Apple's system-provided Tk, because it was fixed already. :-) Now that > Mavericks is out he was able to provide the patch for upstream use. > > After installing Mavericks and testing with my existing build of Tk, I > did see the behavior. After doing a fresh checkout of trunk and 8.5, > applying the patch, and rebuilding Tcl/Tk, I don't see any of the > behavior indicated. As a result I believe this patch solves the problem > and I have committed it. > > Thanks again for the report, and thanks to Daniel for the patch. > > Best, > Kevin > > > -- > Kevin Walzer > Code by Kevin/Mobile Code by Kevin > http://www.codebykevin.com > http://www.wtmobilesoftware.com |
From: Alexander S. <a.s...@gm...> - 2013-10-28 07:31:26
Attachments:
smime.p7s
|
Hi Jeff, my first tests are positiv, I'll do some more test this evening. Best Regards, Alex Am 27.10.2013 um 21:57 schrieb Jeff Hobbs <je...@ac...>: > We are building updated ActiveTcl installations for 8.5 and 8.6 with this patch this weekend. If someone wants to test from that, let me know. > > Jeff > > On 2013-10-27, at 1:46 PM, Kevin Walzer <kw...@co...> wrote: > >> Hi Ned, >> >> On 10/23/13, 4:06 PM, Ned Deily wrote: >>> Now that OS 10.9 Mavericks has been released (and at no cost), people are >>> updating to it and running into a problem with Aqua Tk. I ran into the >>> problem with Python's IDLE but it seems to affect other multi-window Tk >>> applications, as it can be reproduced using wish demos. Basically, the >>> problem is that Tk windows don't get automatically redrawn when the mouse is >>> used to change the active window and focus. I've described the details here: >>> >>> https://core.tcl.tk/tk/tktview?name=53f7a1b553 >>> >>> As it affects current versions of ActiveTcl 8.5 and 8.6, I also opened an >>> ActiveTcl issue: >>> >>> http://bugs.activestate.com/show_bug.cgi?id=101210 >>> >>> As noted in the Tk issue above, the Apple-supplied Tk 8.5.9 does not have this >>> problem but that version has other, serious problems that have been fixed in >>> newer versions of Tk 8.5, for example, immediately crashing when typing a >>> composing character in a text field (like option-u for US Extended input >>> methods). Also as noted, one workaround appears to be to force Tcl/Tk to run >>> in 32-bit mode, which may not be possible for some applications if they depend >>> on non-universal libraries. >> >> Daniel Steffen provided me a patch for the issue that he devised during >> the development of Mavericks, and which he applied to Apple's private >> branch of Tk-Cocoa (essentially the old 8.5.9 branch, as you noted in >> the bug report). Daniel's patch is why the behavior isn't visible in >> Apple's system-provided Tk, because it was fixed already. :-) Now that >> Mavericks is out he was able to provide the patch for upstream use. >> >> After installing Mavericks and testing with my existing build of Tk, I >> did see the behavior. After doing a fresh checkout of trunk and 8.5, >> applying the patch, and rebuilding Tcl/Tk, I don't see any of the >> behavior indicated. As a result I believe this patch solves the problem >> and I have committed it. >> >> Thanks again for the report, and thanks to Daniel for the patch. >> >> Best, >> Kevin >> >> >> -- >> Kevin Walzer >> Code by Kevin/Mobile Code by Kevin >> http://www.codebykevin.com >> http://www.wtmobilesoftware.com > > > ------------------------------------------------------------------------------ > October Webinars: Code for Performance > Free Intel webinars can help you accelerate application performance. > Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from > the latest Intel processors and coprocessors. See abstracts and register > > http://pubads.g.doubleclick.net/gampad/clk?id=60135991&iu=/4140/ostg.clktrk > _______________________________________________ > Tcl-mac mailing list > tc...@li... > https://lists.sourceforge.net/lists/listinfo/tcl-mac -- Alexander Schöpe IT-Sachverständiger Soft-Engineering, Support und Consulting Im Haarmannsbusch 125a 44797 Bochum T: +49 (234) 9799566 F: +49 (3212) 1225370 USt-IdNr: DE272452626 |
From: Ned D. <na...@ac...> - 2013-10-28 08:37:41
|
In article <526...@co...>, Kevin Walzer <kw...@co...> wrote: > On 10/23/13, 4:06 PM, Ned Deily wrote: > > Now that OS 10.9 Mavericks has been released (and at no cost), people are > > updating to it and running into a problem with Aqua Tk. I ran into the > > problem with Python's IDLE but it seems to affect other multi-window Tk > > applications, as it can be reproduced using wish demos. Basically, the > > problem is that Tk windows don't get automatically redrawn when the mouse > > is > > used to change the active window and focus. I've described the details > > here: > > > > https://core.tcl.tk/tk/tktview?name=53f7a1b553 > > > > As it affects current versions of ActiveTcl 8.5 and 8.6, I also opened an > > ActiveTcl issue: > > > > http://bugs.activestate.com/show_bug.cgi?id=101210 > > > > As noted in the Tk issue above, the Apple-supplied Tk 8.5.9 does not have > > this > > problem but that version has other, serious problems that have been fixed > > in > > newer versions of Tk 8.5, for example, immediately crashing when typing a > > composing character in a text field (like option-u for US Extended input > > methods). Also as noted, one workaround appears to be to force Tcl/Tk to > > run > > in 32-bit mode, which may not be possible for some applications if they > > depend > > on non-universal libraries. > > Daniel Steffen provided me a patch for the issue that he devised during > the development of Mavericks, and which he applied to Apple's private > branch of Tk-Cocoa (essentially the old 8.5.9 branch, as you noted in > the bug report). Daniel's patch is why the behavior isn't visible in > Apple's system-provided Tk, because it was fixed already. :-) Now that > Mavericks is out he was able to provide the patch for upstream use. > > After installing Mavericks and testing with my existing build of Tk, I > did see the behavior. After doing a fresh checkout of trunk and 8.5, > applying the patch, and rebuilding Tcl/Tk, I don't see any of the > behavior indicated. As a result I believe this patch solves the problem > and I have committed it. Thanks for tracking down the problem, Kevin! The applied patch does indeed solve the problems seen with Python's IDLE on 10.9. As described in Issue19373, the Python project has just issued maintenance release candidates, 3.3.3rc1 and 2.7.6rc1, for the current Python 3 and 2 releases. The python.org OS X 64-bit binary installers for these releases now include a built-in version of Tcl/Tk 8.5 so that users no longer have to install third-part Tcl/Tk releases to avoid the critical problems in the system Tcl/Tk 8.5 shipped by Apple in recent OS X releases. The fix arrived after the "rc1" installers were created but "rc1_rev1" installers that include this fix should be available on the 3.3.3 and 2.7.6 download pages soon. http://www.python.org/download/releases/3.3.3/ http://www.python.org/download/releases/2.7.6/ http://bugs.python.org/issue19373 -- Ned Deily, na...@ac... |
From: Russell E. O. <ro...@uw...> - 2013-10-28 20:45:33
|
In article <nad...@ne...>, Ned Deily <na...@ac...> wrote: >... > Thanks for tracking down the problem, Kevin! The applied patch does indeed > solve the problems seen with Python's IDLE on 10.9. > > As described in Issue19373, the Python project has just issued maintenance > release candidates, 3.3.3rc1 and 2.7.6rc1, for the current Python 3 and 2 > releases. The python.org OS X 64-bit binary installers for these releases > now > include a built-in version of Tcl/Tk 8.5 so that users no longer have to > install third-part Tcl/Tk releases to avoid the critical problems in the > system Tcl/Tk 8.5 shipped by Apple in recent OS X releases. The fix arrived > after the "rc1" installers were created but "rc1_rev1" installers that > include > this fix should be available on the 3.3.3 and 2.7.6 download pages soon. > > http://www.python.org/download/releases/3.3.3/ > http://www.python.org/download/releases/2.7.6/ > http://bugs.python.org/issue19373 Thank you both. Do you have any idea how to use my own Tcl/Tk instead of the one provided? I'll need to do this if the crashing bug I recently reported (caused by resizing a font in the option database) has not been fixed (and I'm pretty sure it hasn't; it's too recent and this Mavericks problem was too important). I'm really uneasy about standard Mac Python including Tcl/Tk because Aqua Tcl/Tk has so many bugs. The best version of Tcl/Tk for one application may not be best version for another user. For instance in this case I'll probably need to continue to use 8.5.11 for quite some time, and just force the application into 32-bit mode for compatibility with MacOS X 10.9. -- Russell |
From: Ned D. <na...@ac...> - 2013-10-28 23:28:29
|
In article <row...@ne...>, "Russell E. Owen" <ro...@uw...> wrote: > In article > <nad...@ne...>, > Ned Deily <na...@ac...> wrote: > >... > > Thanks for tracking down the problem, Kevin! The applied patch does indeed > > solve the problems seen with Python's IDLE on 10.9. > > > > As described in Issue19373, the Python project has just issued maintenance > > release candidates, 3.3.3rc1 and 2.7.6rc1, for the current Python 3 and 2 > > releases. The python.org OS X 64-bit binary installers for these releases > > now > > include a built-in version of Tcl/Tk 8.5 so that users no longer have to > > install third-part Tcl/Tk releases to avoid the critical problems in the > > system Tcl/Tk 8.5 shipped by Apple in recent OS X releases. The fix > > arrived > > after the "rc1" installers were created but "rc1_rev1" installers that > > include > > this fix should be available on the 3.3.3 and 2.7.6 download pages soon. > > > > http://www.python.org/download/releases/3.3.3/ > > http://www.python.org/download/releases/2.7.6/ > > http://bugs.python.org/issue19373 > > Thank you both. > > Do you have any idea how to use my own Tcl/Tk instead of the one > provided? Yes, i do. But I was hoping to not have to tell. :=) > I'll need to do this if the crashing bug I recently reported > (caused by resizing a font in the option database) has not been fixed > (and I'm pretty sure it hasn't; it's too recent and this Mavericks > problem was too important). > > I'm really uneasy about standard Mac Python including Tcl/Tk because > Aqua Tcl/Tk has so many bugs. The best version of Tcl/Tk for one > application may not be best version for another user. For instance in > this case I'll probably need to continue to use 8.5.11 for quite some > time, and just force the application into 32-bit mode for compatibility > with MacOS X 10.9. I was concerned as well about use cases such as yours. The primary focus of the python.org installers, IMO, is to make it easy for inexperienced users to get going with Python quickly, e.g. "batteries included". But that doesn't mean we should preclude use by more sophisticated users. And, one of the "sophisticated" use cases for the python.org installers is to build Python applications that run on multiple versions of OS X. So, although at the moment the option is not highlighted and there is not a easy to user interface and it is subject to change, it is possible to continue to use to Tkinter with a third-party Tcl/Tk in /Library/Frameworks or the Apple-supplied Tcl/Tk in /System/Frameworks. The new python.org 64-bit installers actually ship with two versions of the _tkinter.so extension module, one that links with the new built-in Tcl/Tk libraries and another that continues to link as before with /Library/Frameworks Tcl and Tk frameworks (falling back to /System/Library/Frameworks). The first is the default installed version. There are two options available. If you want to change your Python installation to use the latter _tkinter.so for all users, you can follow the instructions in the source tree here: http://hg.python.org/cpython/file/v3.3.3rc1/Mac/BuildScript/README.txt http://hg.python.org/cpython/file/v2.7.6rc1/Mac/BuildScript/README.txt The idea is to just copy the desired _tkinter.so into lib-dynload: sudo bash cd /Library/Frameworks/Python.framework/Versions/3.3 cd ./lib/python3.3 cp -p ./lib-tkinter/library/_tkinter.so ./lib-dynload exit It is also possible to modify the Python module search path to find the other _tkinter first. This allows changing _tkinter's dynamically without requiring administrator access. An example: # To use the "library" tkinter -> /Library/Frameworks PYTHONPATH=/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/lib- tkinter/library /usr/local/bin/idle2.7 # To return to the "builtin" tkinter -> using Python's built-in copy of Tcl/Tk PYTHONPATH=/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/lib- tkinter/builtin /usr/local/bin/idle2.7 I have to admit that I came up at the last minute with the idea of putting the _tkinter's into that kind of directory structure so that you could use PYTHONPATH or manipulate sys.path if you know the right paths. I didn't have time to try to make it more elegant or properly document it prior to rc1 for 3.3.3 and 2.7.6. I would consider it "experimental". I'm open to suggestions on other approaches but they are not likely to make it into the final releases of 3.3.3 and 2.7.6 at this point. (Note, that the current 3.4.0a4 has an earlier implementation and doesn't work quite like this. The current tip of the default trunk does and, unless changed, will be in 3.4.0b1.) -- Ned Deily, na...@ac... |
From: Ned D. <na...@ac...> - 2013-11-12 07:33:30
|
In article <nad...@ne...>, Ned Deily <nad...@pu...> wrote: > In article > <row...@pu...>, > "Russell E. Owen" <row...@pu...> wrote: > > I'm really uneasy about standard Mac Python including Tcl/Tk because > > Aqua Tcl/Tk has so many bugs. The best version of Tcl/Tk for one > > application may not be best version for another user. For instance in > > this case I'll probably need to continue to use 8.5.11 for quite some > > time, and just force the application into 32-bit mode for compatibility > > with MacOS X 10.9. > I was concerned as well about use cases such as yours. The primary focus of > the python.org installers, IMO, is to make it easy for inexperienced users to > get going with Python quickly, e.g. "batteries included". An update: after Python 2.7.6rc1 and 3.3.3rc1 received some exposure, a major compatibility problem was discovered with the built-in Tcl/Tk 8.5 implementation in the python.org installer. Several important third-part packages that depend on the installer also depend on building and/or dynamically linking with Tcl and Tk frameworks in /Library. These include PIL/Pillow and Matplotlib. Given the urgency of getting updated installers out to support 10.9 Mavericks, we decided to pull the built-in Tcl/Tk support out of 2.7.6 (final) and 3.3.3 (as of rc2). So, these releases work the way previous releases have wrt which Tcl and Tk libraries are used. What to do for 3.4.0 is under consideration, but the built-in implementation will definitely change for 3.4.0b1. Thanks for your input, Russell. -- Ned Deily, na...@ac... |
From: Russell E. O. <ro...@uw...> - 2013-11-12 18:32:32
|
In article <nad...@ne...>, Ned Deily <na...@ac...> wrote: > In article > <nad...@ne...>, > Ned Deily <nad...@pu...> wrote: > > In article > > <rowen-938790.13450928102013-2AO0Uh8ossnZ+VzJOa5vwg-XMD5yJDbdMReXY1tMh2IBg@p > > ublic.gmane.org>, > > "Russell E. Owen" > > <row...@pu...> wrote: > > > I'm really uneasy about standard Mac Python including Tcl/Tk because > > > Aqua Tcl/Tk has so many bugs. The best version of Tcl/Tk for one > > > application may not be best version for another user. For instance in > > > this case I'll probably need to continue to use 8.5.11 for quite some > > > time, and just force the application into 32-bit mode for compatibility > > > with MacOS X 10.9. > > I was concerned as well about use cases such as yours. The primary focus > > of > > the python.org installers, IMO, is to make it easy for inexperienced users > > to > > get going with Python quickly, e.g. "batteries included". > > An update: after Python 2.7.6rc1 and 3.3.3rc1 received some exposure, a major > compatibility problem was discovered with the built-in Tcl/Tk 8.5 > implementation in the python.org installer. Several important third-part > packages that depend on the installer also depend on building and/or > dynamically linking with Tcl and Tk frameworks in /Library. These include > PIL/Pillow and Matplotlib. Given the urgency of getting updated installers > out to support 10.9 Mavericks, we decided to pull the built-in Tcl/Tk support > out of 2.7.6 (final) and 3.3.3 (as of rc2). So, these releases work the way > previous releases have wrt which Tcl and Tk libraries are used. What to do > for 3.4.0 is under consideration, but the built-in implementation will > definitely change for 3.4.0b1. Thanks for your input, Russell. Good catch. I had not thought that through. I had assumed your python installer put Tcl/Tk into its usual spot in /Library/Frameworks, in which case it would not affect how one built matplotlib and PIL. But that would replace any existing Tcl/Tk, which could be really nasty for users (especially since ActiveState does not serve older versions!). If you do decide to got that route, I hope the installer will ask before deleting an existing Tcl/Tk 8.5. I'm sorry this is turning out to be such a headache. Much as I like my Mac, I sure hate the hassles in building unix-ish software for it. -- Russell P.S. your instructions for using a separate Tcl/Tk with a Python that includes its own Tcl/Tk seemed clear and simple enough. That addressed my concern that you quote above. |