Tk Library Source Code

When running the testsuite for 'math' I get three
failures in the math::fuzzy testsuite. Attached the log
for the whole math package.

Logged In: YES 
user_id=400048

I have been too hasty - I apologize. I did not do a cvs
update. Curiously enough I also did not get any conflicts
... 

I have not had the luxury to work with CVS these past few
years (we switched to another product that I still can not
feel at ease with), so I have to relearn the proper
etiquette.

Logged In: YES 
user_id=75003

Committed re-applied fixes. Verified that the changes fix this 
bug too. Done.

Logged In: YES 
user_id=75003

Thank you. With a weeping eye. Did you run a cvs update 
before your changes ? Or did you get a message about 
conflicts when you tried to commit, and then copied the new 
version of the one provided by the CVS ?

Because the improved version is based on the code you 
originally submitted, and thus overwrites the bugfixes and 
cleanups I did. Especially the bugfix to make the file operable 
overall as part of the tcllib testsuite.

Ok, water under the bridge. I will now re-apply my changes to 
your improved version to get these fixes back. However I do 
ask you to make sure that your development is either based 
on the code we have in the tcllib CVS, or to make sure that 
your updates are correctly merged instead of simply 
overwriting anything which was done with what is essentially 
an old version with different changes.

Logged In: YES 
user_id=400048

I have committed an improved version

Logged In: YES 
user_id=400048

Okay, that is a very reasonable (and probably the only good)
solution. I will repair this and submit a new test suite

Logged In: YES 
user_id=75003

No. Tests never fail(in the absence of bugs).
If a test drives the functionality intentionally into an error state 
then the test has to check that the error state has occured, 
and report that as a sucess. And if the intended error state 
did not occur, then that is a failure of the functionality, and 
that is a failure of the test.

Example (taken from struct, graph)

test graph-0.1 {graph errors} {
    graph mygraph
    catch {graph mygraph} msg
    mygraph destroy
    set msg
} "command \"mygraph\" already exists, unable to create 
graph"

We intentionally call commands in an illegal way, catch the 
error message produced by the graph commands and check 
that this message is the correct one. If it is the test is 
sucessful. If the message is not right, then the test failed, as 
something in the functionality is not as we expected it.

Logged In: YES 
user_id=400048

These tests are actually meant to fail (see the
description). They fail because in the version of tcltest
that was used during development of the package, it was not
possible to set up custom comparison routines.

(It led to TIP #85 and the corresponding new version of
tcltest. Perhaps it is time to update the test file.)

File Added - 47682: LOG