[Tcl-bugs] [ tktoolkit-Bugs-741179 ] text-15.* fails to clear font

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

Bugs item #741179, was opened at 2003-05-21 15:12
Message generated for change (Settings changed) made by vincentdarley
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=112997&aid=741179&group_id=12997

Category: 18. [text]
Group: 8.5a0
>Status: Closed
>Resolution: Fixed
Priority: 5
Submitted By: Don Porter (dgp)
Assigned to: Vince Darley (vincentdarley)
Summary: text-15.* fails to clear font

Initial Comment:

$ make test TESTFLAGS="-file text.test -match text-15.*"
TCL_LIBRARY=/local/src/tcl/library; export TCL_LIBRARY; \
TK_LIBRARY=/local/src/tk/library; export TK_LIBRARY; \
./tktest /local/src/tk/unix/../tests/all.tcl -geometry
+0+0 \
-file text.test -match text-15.* 
Tests running in interp:  /local/src/tk/solaris/tktest
Tests located in:  /local/src/tk/tests
Tests running in:  /local/src/tk/solaris
Temporary files stored in /local/src/tk/solaris
Test files sourced into current interpreter
Running tests that match:  text-15.*
Skipping test files that match:  l.*.test
Only running test files that match:  text.test
Tests began at Wed May 21 11:11:07 AM EDT 2003
text.test

Tests ended at Wed May 21 11:11:08 AM EDT 2003
all.tcl:        Total   358     Passed  2       Skipped
356     Failed  0
Sourced 2 Test Files.
Font Courier -12 still in cache.


Note the final line.

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

Comment By: Don Porter (dgp)
Date: 2003-05-22 15:58

Message:
Logged In: YES 
user_id=80530


That patch does make the
"still in cache" message go away, yes.

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

Comment By: Vince Darley (vincentdarley)
Date: 2003-05-21 23:31

Message:
Logged In: YES 
user_id=32170

Yup, one unfortunate side-effect of the new option-
freeing code (Obj-based) is that it actually requires a 
tkwin to operate on.  The old stuff didn't necessarily, 
and it is these changes that are causing this problem, in 
conjunction with a design which overloaded 'textPtr-
>tkwin == NULL' as meaning the widget was being 
deleted.

Anyway, please try the attached patch.


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

Comment By: Don Porter (dgp)
Date: 2003-05-21 20:16

Message:
Logged In: YES 
user_id=80530

Yes, bugginess in widget destruction is
what I'd expect.  Several Tk widgets
have been through similar troubles.

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

Comment By: Vince Darley (vincentdarley)
Date: 2003-05-21 19:09

Message:
Logged In: YES 
user_id=32170

I imagine the problem is in lines 1346 onwards of 
tkText.c where the widget is destroyed.  If the tkwin 
structure is already null'd out, then the config options 
will not be freed (and therefore the font will not be 
freed).  It looks as if this can happen when called from 
TextCmdDeletedProc, i.e. if the widget command is 
deleted rather than the window being destroyed.

Therefore removing 'textPtr->tkwin = NULL' from 
TextCmdDeletedProc ought to fix this problem, but may 
well create other problems, particularly around double-
deletion of the widget.

Unfortunately, I can't seem to reproduce the problem on 
Windows, so am having a hard time looking into it.


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

Comment By: Don Porter (dgp)
Date: 2003-05-21 16:52

Message:
Logged In: YES 
user_id=80530


when I revert to the Tk HEAD
of 2003-05-17, the problem
goes away.

The message comes from 
TkFontPkgFree in tkFont.c.

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

Comment By: Vince Darley (vincentdarley)
Date: 2003-05-21 16:18

Message:
Logged In: YES 
user_id=32170

Do you know if this bug was in the old text widget? Also 
can you tell me where that 'still in cache' message 
comes from....

cheers, Vince.


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

Comment By: Donal K. Fellows (dkf)
Date: 2003-05-21 15:22

Message:
Logged In: YES 
user_id=79902

I've not touched that bit at all...

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

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=112997&aid=741179&group_id=12997