view release on metacpan or  search on metacpan

Todo  view on Meta::CPAN

> Linux:
> http://www.nntp.perl.org/group/perl.daily-build.reports/43396    29609
> http://www.nntp.perl.org/group/perl.daily-build.reports/43381    29604
> 																														     
> AIX:																												     
> http://www.nntp.perl.org/group/perl.daily-build.reports/43428    29609
> http://www.nntp.perl.org/group/perl.daily-build.reports/43395    29604
> 																														     
> OpenBSD:																										     
> http://www.nntp.perl.org/group/perl.daily-build.reports/43414    29609
> http://www.nntp.perl.org/group/perl.daily-build.reports/43374    29599

	That's something different. So I reported the 29er to p5p. It's since
	been fixed in 29639.

2006-12-28  Andreas J. Koenig  <andreas.koenig.7os6VVqR@franz.ak.mind.de>

	* Bug in SQLite: if I type "install TAB" I only get completion for the
	modules that have already been loaded:

cpan[5]> install             
Acme::Meta      Danga::Socket   POSIX           Sys::Syscall    Time::HiRes
DBD::mysql      IO::Poll        Socket          Test::More      fields

	Reported to and FIXED by Randy.

2006-12-27  Andreas J. Koenig  <andreas.koenig.7os6VVqR@franz.ak.mind.de>

	* #24114 by Steve says that we do not survive without YAML!

	Trying to fix with overloaded Exception classes leads to more changes
	than I feel comfortable with without writing new tests. So we need tests
	that test missing YAML and YAML::Syck. Maybe Test::Without::Module might
	help?

	No. The tricky way that CPAN.pm is using to load modules is not suited
	for T:W:M. I have tried it and I see YAML gets loaded without
	hesitation. But I recall we have dontload_list. Very nice.

	IMPROVED in SVK rev. 1639 and added tests in 31sessions.t since then.

	* document the major change to the force pragma and find tests that
	enforce that

	* document how test repete is supposed to work and find tests that
	enforce that

2006-12-26  Andreas J. Koenig  <andreas.koenig.7os6VVqR@franz.ak.mind.de>

	* Paul Johnson has already fixed #24067 in his repository. Unfortunately
	he sends me a link to svnweb instead of the repository itself. Svnweb
	itself neither gives machine readable output nor does it provide a link
	to the repository.

	Fortunately one can guess it: svn ls http://pjcj.sytes.net/svn/Devel-Cover

	Repository version passes all tests.

2006-12-25  Andreas J. Koenig  <andreas.koenig.7os6VVqR@franz.ak.mind.de>

	* I don't know how to prevent SQLite dieing due to locked database.
	Because it happens such that a batch process reads the database and an
	interactive process tries to write the database. The latter does not
	know about the other. Both know nothing about each other beforehand.
	Then they clash and I cannot find out.

	FIXED (at least improved) by Randy

2006-12-24  Andreas J. Koenig  <andreas.koenig.7os6VVqR@franz.ak.mind.de>

	* Todo: create a second shell test, say 31force.t that starts from
	scratch and studies the variants without having to wait for many minutes

	It would be good if we had a test script that allows us to visit
	multiple sessions. It would not really need Expect.pm! We make short
	sessions instead. 31sessions.t is it. Test::Base or what? No, it doesn't
	convince me but I don't know why.

	* after SVK rev. 1604 doing a megainstall from scratch a second time to
	study the differences tomorrow or in the late evening.

	Later... looks good. But crazy that we find both DBD::mysql 4.00 and
	3.0008 in the failed departement (Update 2006-12-26 akoenig : reported
	as http://rt.cpan.org/Ticket/Display.html?id=24096). And the second
	anomaly that .../svn/tk/. succeeded at 17:28 but not at 11:45. Not much
	for 5100 diff chunks.

	* Rethinking once again 'force test'. Right now I see "Has already been
	tested successfully" within one session but not when I start a session.

	This needs to change because our big force redesign should let us say
	'force test thing' when we really want just exactly that and should save
	us the time when we really know this has been tested OK.

	I think this happens in SVK rev.1604

	* Buglet: if I do a 'force get thing' then I get a new build_dir for the
	thing. If I had done a 'test thing' before then the old build_dir is in
	the is_tested hash/set/database and survives in the paths prepended to
	PERL5LIB

	FIXED in SVK rev. 1589

	But wait! If we have many variants of one distro in PERL5LIB, then all
	hell breaks loose because we possibly do not get the right one as the
	first. We need to sort by something. DONE in upcoming SVK rev. 1604.

	Oh, and we need to filter out the installed ones. So we really must tie
	the distros and the is_tested stuff together. Makes no sense to prepend
	test-directory-X to PERL5LIB when X is installed. FIXED for the
	make_install attribute: this was lost because there was an early return
	on success in CPAN::Distribution::install().

2006-12-23  Andreas J. Koenig  <andreas.koenig.7os6VVqR@franz.ak.mind.de>

	* Remember to see after the Devel::Cover module

	http://rt.cpan.org/Ticket/Display.html?id=24067

	It seems to me that the CC to p5p did not come through. Need to re-check
	and forward it myself. But give it time to bite its way through.