[Tcllib-devel] Update regarding testsuite results

[Andreas K. <and...@Ac...>]

After the updates done to the CVS in the last two days the summary results
look better.

They are actually very very good, but this becomes only clear when reading
the full reports about which tests failed and why. I have done this already
and have annotated the summary with my findings, for easy reading. No need
for anyone else to get and dig through the full reports.


System Architecture : linux-ix86

Shell    Module   Total Passed Skipped Failed
-----    ------   ----- ------ ------- ------
tclsh8.5 asn      216   204    0       12		[1]
-----    ------   ----- ------ ------- ------
tclsh8.5 counter  16    13     1       2		[2]
-----    ------   ----- ------ ------- ------
tclsh8.4 docstrip 7     6      0       1		[4]
tclsh8.5 docstrip 7     6      0       1		[4]
-----    ------   ----- ------ ------- ------
tclsh8.5 exif     1     0      0       1		[2]
-----    ------   ----- ------ ------- ------
tclsh8.5 math     1236  1235   0       1		[2]
-----    ------   ----- ------ ------- ------
tclsh8.5 profiler 20    14     5       1		[3]
-----    ------   ----- ------ ------- ------
tclsh8.5 units    97    96     0       1		[2]
-----    ------   ----- ------ ------- ------
Shell    Module   Total Passed Skipped Failed

[1] The asn testsuite actually exposes a bug in Tcl 8.5 and
    how it is processing numbers. An error has been raised
    for this at the Tcl tracke (IIRC).

	Nothing has to be done by Tcllib.

[2] The treatment of floats has been changed in Tcl 8.5
    (tcl_precision default IIRC). This causes some float
    results to change between 8.4 and 8.5. It will change
    further with Kevin's bignum/floats work.

	I see no reason to try and track these changes until
	that area of Tcl 8.5 has stabilized. I.e. nothing has
	to be done by Tcllib maintainers.

[3] In Tcl 8.5 some clock commands have been implemented as
    procedures, and are thus now reported by the profiler.
    The testsuite does not account for this. I am not sure
    if we should simply update the testsuite for this, if
    some type of exclusion in the profiler itself is warranted.

[4] This is a bug in the testsuite of docstrip and how it tries
    to access some files in the tested module.

    I am willing to let this one slide, although fixing this is
    trivial. It would IMHO be nicer to fix it, as a bit of polish.

So, while the summary is not looking like it, the testsuite is now basically
in order, save the docstrip nit.


Question is, will it look that good for other platforms as well ?

Please get the testenv and run it on your platforms as well. The more, the
merrier. And the better covered we are.

--
	Andreas Kupries <and...@Ac...>
	Developer @ http://www.ActiveState.com, a division of Sophos
	Tel: +1 604 484 6491