Test::Unit

2001-02-17 Thread schwern
I was poking around on CPAN and noticed this rather complete alternative to Pod::Tests for embedding tests in code, a bit closer to what Barrie was discussing. I've invited the author onto the list and hopefully he'll say a few words. I'd also like people to poke around with it, see what its all

Putting Test::Harness back on CPAN?

2001-02-17 Thread schwern
Part of the reason I'm going all this work on Test::Harness is because I'm going to need the new features for my next job (altruism isn't dead, its just lying bleeding on the floor). As such, I need the new Test::Harness *without* having to use bleedperl. Judging from the discussion on perl-qa

[PATCH t/TEST lib/Test/Harness.pm] Adding todo tests

2001-02-17 Thread schwern
A few weeks ago I brought up the idea of unifying the format of todo tests and skip tests: http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2001-01/msg00883.html Well, here it is. This is a patch to both t/TEST and Test::Harness so they honor this style of test output: not ok 13 #

Re: Why t/TEST and not Test::Harness?

2001-02-17 Thread schwern
On Sun, Feb 18, 2001 at 04:44:05AM +, Simon Cozens wrote: > On Sat, Feb 17, 2001 at 09:46:31PM -0500, [EMAIL PROTECTED] wrote: > > Why is t/TEST anything more than a thin wrapper around Test::Harness? > > Because of potential fragility. If Perl isn't entirely together when > you're running th

Re: Why t/TEST and not Test::Harness?

2001-02-17 Thread Simon Cozens
On Sat, Feb 17, 2001 at 09:46:31PM -0500, [EMAIL PROTECTED] wrote: > Why is t/TEST anything more than a thin wrapper around Test::Harness? Because of potential fragility. If Perl isn't entirely together when you're running the tests, the last thing you want to do is load up Test::Harness. t/TEST

Re: Why t/TEST and not Test::Harness?

2001-02-17 Thread schwern
On Sat, Feb 17, 2001 at 09:08:15PM -0600, Jarkko Hietaniemi wrote: > I think it was supposed to be TEST with -Mutf8, basically. > Why it was a separate script? Beats me. Well, its not anymore. I've added a -utf8 flag to t/TEST and elminated t/UTEST. make utest still works. Here's a patch of t

Re: Why t/TEST and not Test::Harness?

2001-02-17 Thread schwern
And what the hell is t/UTEST? Looks like its for unicode. Ok, fine, but I don't like the code duplication. In fact, it already looks like the two are falling out of sync. Case in point, "make utest" still breaks under the 'ok 1 - name' style. It was never patched along with TEST. Additionall

Why t/TEST and not Test::Harness?

2001-02-17 Thread schwern
Why is t/TEST anything more than a thin wrapper around Test::Harness? It contains a partial reimplementation, but is missing some very useful features, such as skip and todo tests. I particularly miss todo tests. There's alot of little bugs which people mention and then are forgotten, I'd like t

Re: Reorganizing t/

2001-02-17 Thread schwern
On Sat, Feb 17, 2001 at 07:49:37PM +, [EMAIL PROTECTED] wrote: > >A simpler idea would to alter t/TEST so that it runs the tests as > >"perl -I../lib". This isn't *quite* the same as @INC = '../lib', but > >it should be close enough. > > It isn't. More than once we have things "pass" for _us

Re: Reorganizing t/

2001-02-17 Thread nick
<[EMAIL PROTECTED]> writes: > >A simpler idea would to alter t/TEST so that it runs the tests as >"perl -I../lib". This isn't *quite* the same as @INC = '../lib', but >it should be close enough. It isn't. More than once we have things "pass" for _us_ 'cos we have perl already and test finds /us

Re: Reorganizing t/

2001-02-17 Thread schwern
On Sat, Feb 17, 2001 at 10:55:29AM -0600, Jarkko Hietaniemi wrote: > On Sat, Feb 17, 2001 at 10:22:13AM -0600, Jarkko Hietaniemi wrote: > > > One additional point to consider and solve is the BEGIN 'preamble' we > > > now add to each test to make certain we are in a known directory, and > > > more

Re: Reorganizing t/

2001-02-17 Thread Simon Cozens
On Sat, Feb 17, 2001 at 10:05:56AM -0800, Russ Allbery wrote: > As one of the people with modules both on CPAN and in core, I'd be happy > to put such a pre-amble on all of my tests in my module distribution if it > didn't interfere with people running make test from the module > distribution. I

Re: Reorganizing t/

2001-02-17 Thread Russ Allbery
Jarkko Hietaniemi <[EMAIL PROTECTED]> writes: > Looks good to me. > One additional point to consider and solve is the BEGIN 'preamble' we > now add to each test to make certain we are in a known directory, and > more importantly, that we are running the tests using *the* lib/ of the > Perl we ar

Reorganizing t/

2001-02-17 Thread schwern
Alot of the modules in the core have (or had) double lives on CPAN. Alot of the tests are enormous. And t/ is, in general, something of a mess. Part of the problem is its fairly flat directory structure. t/ is broken up into only a few subdirectories: base, cmd, comp, io, lib, op, pod, pragma.

Preliminary test coverage analysis

2001-02-17 Thread schwern
I've wedged coverage analysis into Test::Harness (not quite ready for release yet) and ran it over Perl's core test suite to see how well the core modules are covered. I'm not 100% sure about this data, Devel::Coverage needs alot of work and reported alot of false negatives, but I think I can mak