Re: [Cgi-session-user] [Fwd: FAIL CGI-Session-4.29_1 MSWin32-x86-multi-thread 5.1]

[Ron S. <ro...@sa...>]

Hi Mark

> Ron Savage,
> 
> Would you be able to look at the 4.29_1 test failures? This one appears 

I've renamed t/bug21981.t to t/bug21981.todo.

It looks like this test was an idea without sufficient planning to
support it. Some versions of Perl don't allow 'use encoding;' by
default, and other versions die on sprintf. If (s)printf has to be
re-thought, then the problem is far from fixed.

> to be UTF related. I haven't looked at the others.

Details here:

http://testers.cpan.org/show/CGI-Session.html

Basically they fall into 2 categories:

(1) Illegal seek

(2) perl Build.PL Build

besides the uft8 problem.

Discussion:

(1)
The Illegal seek seems to only occur in 5.10.0 when DB_File is used
(t/api3_db_file.t). I don't get this error with 5.8.8, and the only
5.10.0 I have is Strawberry Perl, which does not ship with DB_File (and
on which DB_File cannot be installed because of a missing db.h).

The only use of seek in CGI::Session is in CGI::Session::ID::incr, so
I'm wondering if tighter code in 5.10.0 is revealling a bug in DB_File.

As you can see, I'm guessing.

(2)
This obviously relates to me adding Build.PL to the mix, although why
Build.PL is being run with a Build option is beyond me. On one machine,
there was even an extra msg that Module::Build was not installed. It's
disappointing, but perhaps Build.PL should be deleted? Thoughts?

-- 
Ron Savage
ro...@sa...
http://savage.net.au/index.html