tcl 8.4.4

Jacob Levy July 19, 2003: Please report here on your testing of Tcl 8.4.4. It's available by anon CVS now in the CVS.

To get the source code:

  cvs -d:pserver:[email protected]:/cvsroot/tcl login
  cvs -z3 -d:pserver:[email protected]:/cvsroot/tcl co -r core-8-4-4 tcl

Similarly for Tk:

  cvs -d:pserver:[email protected]:/cvsroot/tktoolkit login
  cvs -z3 -d:pserver:[email protected]:/cvsroot/tktoolkit co -r core-8-4-4 tk


  % ./configure --prefix=<whereyouwantitinstalled>
  % make
  % make test
  % make install
  % <any other tests e.g. test your own extensions>

Jacob Levy July 19, 2003: Tested the Tcl side of things on RH 7.3 gcc 2.96, all clean. Cannot run 'make test' on Tk because I have a display-less machine (only ssh-accessible). Tk did build and install cleanly.

TV I make-ed the previous version only, on rh8+, download from web browser, then uncompress and untar:

 tar -xzf tcl*gz

and the same for tk. Same as above, but need tcl install first (not just make) before tk, though I guess some environment var could be set. All clean, too. I used a non-superuser account, which wasn't a problem except for standard install places, has to be root.

Try XFree86, the currently cygwin supplied X server for windows, runs smooth enough. Unless you have another reason things can just be ssh, that nicely can give you the wish window on a windows machine, running on linux.

MC 19 July 2003: I'm going to test building Tcl 8.4.4 with all the various permutations of ./configure --options on Debian Linux, RH 6.0, Tru64 and OpenBSD. Here's the Tcl script I'm using to generate a shell script to do all the compiles and create a log with all the configure, compile and test output. (It'll need a few tweaks for Tru64, i.e., cc -V instead of gcc -v, etc.)

 # -*- tcl -*- \
 exec tclsh $0 ${1+"$@"}

 proc sign var {
     upvar 1 $var thing
     if {$thing == "enable"} {
         return +$var
     } else {
         return -$var

 puts "#!/bin/sh"
 puts "cd ~/testbuild"

 set counter 0

 foreach threads {enable disable} {
     foreach langinfo {enable disable} {
         foreach shared {enable disable} {
             foreach load {enable disable} {
                 foreach symbols {enable disable} {
                     set builddir build[format %02d [incr counter]]
                     append builddir [sign threads]
                     append builddir [sign langinfo]
                     append builddir [sign shared]
                     append builddir [sign load]
                     append builddir [sign symbols]

                     set logfile ~/testbuild/$builddir/build.log

                     puts "# $logfile" 
                     puts "if \[ ! -e $builddir \]; then mkdir $builddir; else rm -rf $builddir/tcl/; fi"
                     puts "cd $builddir"
                     puts "tar xvfz ../tcl8.4.4-cvs-head.tgz"
                     puts "cd tcl/unix/"
                     puts "echo `date` $logfile | tee $logfile"
                     puts "uname -a | tee --append $logfile"
                     puts "gcc -v 2>&1 | tee --append $logfile"
                     puts "./configure -$threads-threads -$langinfo-langinfo -$shared-shared -$load-load -$symbols-symbols 2>&1 | tee --append $logfile"
                     puts "make 2>&1 | tee --append $logfile"
                     puts "make test 2>&1 | tee --append $logfile"
                     puts "cd ~/testbuild"
                     puts -nonewline \n

SRIV Redhat 9.0, tcl8.4.4 make and make test ran with 0 errors. Win2k MinGW tcl8.4.4 make and make test ran with 19 errors. The errors were related to file globbing and normalization, and occur in 8.4.3 as well, probably due to the way the MinGW environment modifys the paths.

Jul 20 2003: Linux-ARM, HP skiffcluster, make test completed with 10 errors, 2 in cmdAH.test while doing atime/mtime tests, due to no access to /tmp. The other 8 were in fCmd.test, while copying/renaming files, again, probably due to no /tmp.

AF had some test failures with vc7 on win xp.

 Tests ended at Sat Jul 19 1:43:08 PM Pacific Daylight Time 2003
 all.tcl:        Total   10503   Passed  9749    Skipped 690     Failed  64
 Sourced 129 Test Files.
 Files with failing tests: httpold.test io.test registry.test winFCmd.test winPipe.test

the full log is at

 Tests ended at Sat Jul 19 2:07:42 PM Pacific Daylight Time 2003
 all.tcl:        Total   8324    Passed  6728    Skipped 1579    Failed  17
 Sourced 182 Test Files.
 Files with failing tests: frame.test grid.test scrollbar.test textImage.test textWind.test winButton.test winWm.test winfo.test

the full log is at

Jacob Levy July 20, 2003: Michael Cleverly did a lot of different compiles on the Tcl 8.4.4 sources, and here are his results:

SO July 21, 2003: A summary of tcltests - built on a win98 box using MSVC++6.0 Std: Tests ended at Sun Jul 20 3:15:20 PM Central Daylight Time 2003 all.tcl: Total 10503 Passed 9764 Skipped 722 Failed 17

Entire test results available at:

I had(have) a problem running the tktests - a piping problem which may be related to to location I built from - will try to work it out this evening

TV '(3 sept 03)' Built tcl/tk 8.4.4 on Linux Redhat 9.0 with all latest updates, including Img package. Smooth and fast compile on a Medion 2.6 G pentium notebook, Img works out of box (after make install). Apart from hard to install driver for Nvidia Gforce FX graphics card, which makes Redhat Linux 9.0 install end in a screen slowly decaying from wrong X server settings, having tricked the nvidia opengl driver onto the sysm and a recompile of it to accomodate 9.0, Img on a bwise canvas is fast business, images the size of the screen can be dragged with hardly any delay. Guess I'll try the sockets, too.