tk: runtime version check fails on Yosemite

[seh@…]

After upgrading to Mac OS Yosemite, I find that wish fails to start (when invoked via gitk) due to a failure in checking the operating system version:

wish[19322]: WARNING: The Gestalt selector gestaltSystemVersion is returning 10.9.0 instead of 10.10.0. Use NSProcessInfo's operatingSystemVersion property to get correct system version number.

Find a Mac OS crash report attached.

[seh@…]

Mac OS crash report for wish

[ned-deily (Ned Deily)]

Actually, I believe the Gestalt version check is non-fatal.

The critical issue that causes the crash is a now faulty version check in Tk's macosx/tkMacOSXInit.c which results in the message:

Mac OS X 10.10000 or later required !

I have opened a Tk bug for this problem: ​https://core.tcl.tk/tk/tktview?name=1d37c3e166. Note that this problem applies to all users of Cocoa Tk (+quartz variant), like Python's IDLE or other Tkinter-based applications. They are all totally broken. However, this issue does not affect the +x11 variant of Tk.

[radaroutthere@…]

Cc Me!

[radaroutthere@…]

Just noting that tickets #45530 and #45532 are also caused by this issue.

Thanks for opening the Tk bug!

[inyeollee]

Cc Me!

[larryv (Lawrence Velázquez)]

Has duplicates #45530 and #45532.

[promac@…]

Cc Me!

[radaroutthere@…]

The official fix for this issue has now been written - ​http://core.tcl.tk/tk/fdiff?v1=64b03ec85cc27fe4de1c03503ffdb2c89f598b3a&v2=7223ddf5d435778396ac2eef3f77f283b5c06394&sbs=1

So either this can be applied as a patch, or the Tk / Tcl ports could be upgraded to the 8.6.3 rc.

[larryv (Lawrence Velázquez)]

Is this a compile-time or run-time problem?

[radaroutthere@…]

If I'm following correctly, the "Mac OS X 10.10000 or later required !" error message is generated at run-time, yes.

[larryv (Lawrence Velázquez)]

I’ve applied the upstream fix in r127151. Could someone verify that 8.6.1_2 works properly?

[davedgd@…]

Replying to radaroutthere@…:

If I'm following correctly, the "Mac OS X 10.10000 or later required !" error message is generated at run-time, yes.

Correct:

[radaroutthere@…]

Yes, the patch is working correctly. Thanks very much!

[promac@…]

Yes. It is working for me too.

Just a curiosity. Why did you use a

# Remove when updated to 8.6.3 (#45444). patchfiles-append fix-yosemite-version-checking.patch

Just to make sure it was the last patch applied?

Thanks.

[larryv (Lawrence Velázquez)]

Replying to promac@…:

Just a curiosity. Why did you use a

# Remove when updated to 8.6.3 (#45444).
patchfiles-append   fix-yosemite-version-checking.patch

To avoid changing other lines in the Portfile. I don’t like when svn blame tells me that line x was last changed in changeset y, only to find out that that all changeset y did was append a backslash or replace tabs with spaces. It’s undeniably unavoidable in many cases, but I try to avoid it when I can. For instance, I commit aesthetic changes before functional ones, so that the functional changesets all show up in svn blame.

In this case, I just added two new lines without touching anything else. When 8.6.3 shows up, these two lines just get deleted. Nice and clean.

[iaindhay@…]

Cc Me!