Tcl 7.4b2 and Tk 4.0b2 Release Announement
January 12, 1995

John Ousterhout
Sun Microsystems, Inc.
john.ousterhout@eng.sun.com

I have just released Tcl 7.4b2 and Tk4.0b2.  Both of these are now
available via FTP from ftp.cs.berkeley.edu (/ucb/tcl/tcl7.4b2.tar.Z
and /ucb/tcl/tk4.0b2.tar.Z) and they should reach the standard Tcl/Tk
mirror sites soon.  There are also patch files to upgrade from the
b1 releases to the b2 releases (tcl7.4b2.patch.Z and tk4.0b2.patch.Z).

These releases should fix the bugs that were causing the most problems
with the b1 releases.  There are many bug reports that I haven't yet
dealt with, so don't be surprised if a bug that you reported isn't
fixed here, but I wanted to get out fixes for the major bugs ASAP.
If I have responded to a bug report of yours and claimed to have fixed
it, but it still doesn't work right in b2, then please let me know.

Here is an extract from Tcl's "changes" file that tells what is fixed
in Tcl7.4b2:

    284. 12/26/94 Fixed "install" target in Makefile (couldn't always
    find install program).
    
    285. 12/26/94 Added strcncasecmp procedure to compat directory.
    
    286. 1/3/95 Fixed all procedure calls to explicitly cast arguments:
    implicit conversions from prototypes (especially integer->double)
    don't work when compiling under non-ANSI compilers.  Tcl is now clean
    under gcc -Wconversion.
    
    287. 1/4/95 Fixed problem in Tcl_ArrayCmd where same name was used for
    both a label and a variable;  caused problems on several older compilers,
    making array command misbehave and causing many errors in Tcl test suite.
    
Here's an extract from Tk's "changes" file that tells what is fixed
in Tk4.0b2:
    
    12/26/94 (bug fix) Removed obsolete demos from Makefile (color, dialog,
    size), fixed "install" target.
    
    1/3/95 (bug fix) Fixed all procedure calls to explicitly cast arguments:
    implicit conversions from prototypes don't work when compiling under
    non-ANSI compilers.  Tk is now clean under gcc -Wconversion.
    
    1/4/95 (bug fix) Used "screenX" without ever setting it in DisplayText
    in tkCanvText.c:  caused tabs in canvas text items to get messed up.
    
    1/4/95 (bug fix) Canvases forgot to register the built-in types if
    Tk_CreateItemType was called before a canvas widget was created.
    
    1/4/95 (bug fixes) Fixed glitches in various text bindings:
	- Up used to do nothing if the cursor was at 2.0.
	- Right used to make the cursor invisible if it was just before
	  the final newline of the text.
	- Control-t didn't conform to Emacs;  made it conform to GNU Emacs.
	- Deleted Control-x binding, since it doesn't conform to anything and
	  is confusing for Emacs users.
    
    1/4/95 (bug fixes) Changed Control-t for entries just as for texts (see
    above) an deleted Control-x for entries (see above).
    
    1/4/95 (bug fix) The packer didn't map slaves unless the master was mapped;
    this could cause slaves to get "lost" so that they weren't mapped until the
    master resized.
    
    1/5/95 (bug fix) Scrollbars weren't executing the proper code the first time
    the mouse entered the widget;  this caused problems if tk_strictMotif was
    set.
    
    1/6/95 (bug fix) Fixed label/button/checkbutton/radiobutton/menubutton
    widgets to allow arbitrary screen distances when specifying -width and
    -height for an image or bitmap (the manual pages already documented this
    but the code didn't implement it).
    
    1/6/95 (new feature) Added very primitive support for input methods,
    as suggested by Martin Forssen.  This should be enough for European
    character sets (Compose key) but it isn't near enough for Asian
    character sets.
    
    1/8/95 (bug fix) Fixed problem in canvas "xview" and "yview" commands
    where divide-by-zero errors could sometimes occur.
    
    1/8/95 (bug fix) New event handler didn't properly handle files for
    which both TK_READABLE and TK_WRITABLE were specified.
    
    1/11/95 (bug fix) Fixed bug with text selections:  was returning count
    too high for data, causing bogus garbage to appear when selection was
    copied.