FAREWELL NOTE
I believe that after epic BSD-battle I leave F::AF in good shape.
What you see on cpan-testers (well, before this release) is
manifestation of APT-methods changing underground. At the time I
believed that I absolutely must know what's happening on other end of
the pipe (and I still hold to this belief). So the test-suite is
doing all terrible things I can implement (admitedly, it's "copy" and
"file" only, everything else is networking). Then APT-methods have
changed.
Updating the test-suite (that's what's de-synchronized) is easy, I
just have to upgrade to something more current. And here comes the
problem. I have no resources to build RedHat for third time (I did it
twice already). And I have to move on. And I won't see Debian for
any foreseeable future. Here we are.
OVERVIEW
File::AptFetch provides API for APT (Advanced Package Tool, that's the
Debian package management system) methods. APT methods aren't
libraries, they are executables (placed, by default, in
/var/lib/apt/methods). So it's not XS (and can't be), F::AF forks.
However one doesn't need fork for each file or session. The IPC is
pipe (and can't be anything else). As you can see there's a limit of
number of methods available simultaneously posed by the maximum number
of open filehandles (APT methods aren't that numerous anyway).
The module is named File::AptFetch because it always fetches. I think
that's possible to develop a method that would bring away files, but,
at present, there's none. And I doubt there would be any.
Further reading:
File::AptFetch
POD of the module. Covers API and provides examples of intended
usage. It's development is kind of dead (unless APT methods API
would change dramatically) -- any additions are in support for
File::AptFetch::Simple.
File::AptFetch::Cookbook
Provides some useful, I believe, findings on each method behaviour
and an overview, of my understanding, how that stuff works. Look,
APT methods API documentation is quite outdated. Each method is
undocumented completely. (bug) My knowledge of APT methods is
very irregular, so is F::AF::C.
File::AptFetch::Simple
That's what you should look into for doing all kinds of stuff.
(bug) It still lacks some functionality.
RUN-TIME REQUIREMENTS
Perl
*v5.8.0* or later. Look, *v5.6.2* is "woody", 20031114. You must
be kidding.
base
Carp
Cwd Any version. Distributed with Perl itself.
IO::Pipe
Any version. It's distributed with Perl itself too. In core
since *v5.3.7*, predates debian, BTW.
List::Util
Required by File::AptFetch::Simple for doing some gibberish.
*v1.03* shall be enough. Distributed with Perl itself.
String::Truncate
Required by File::AptFetch::Simple. *v0.5* looks OK.
version
*v0.77*. CPAN doesn't remember when *v0.50* has been released.
Required version has been upgraded in fruitles attempt to fix some
FTBFSes. Introduced more breakage. Stayed.
BUILD-TIME REQUIREMENTS
File::Temp
POSIX
Test::More
Any version. Should be distributed with Perl itself.
BUILD-TIME CONFIGURATION
*lib_method*
That sets location of APT methods. It's unrequired, it will be
set any time APT configuration is sourced.
"_cache_configuration()" in File::AptFetch and
"_uncache_configuration()" in File::AptFetch have more.
*config_source*
Required. "ARRAY" ref that will be execed.
"_cache_configuration()" in File::AptFetch has more.
*signal*
Required. Names a signal that will be sent to method. "DESTROY"
in File::AptFetch has more.
*timeout*
*tick*
*(v0.0.8)* Required. I've believed, that could stay hardcoded. I
was wrong. "_read()" in File::AptFetch has more.
*(v0.1.6)* These are closely related. From now on, *timeout*
keeps the same semantics as before (it's time to cut-off) but is
handled differently. F::AF object sleeped (literally) for time
set. Now it selects (with magic) in *tick*s. (It's introduced
for sake of F::AF::S.)
Also. Both must be naturals; They're in seconds; *tick* is
supposed to be couple of seconds (default: 5sec); While *timeout*
is supposed to be large enough (default: 2min) in respect to
network introduced delays.
*beat*
*wink*
*(v0.1.9)* *(F::AF::S -- v0.1.4)* Those two are for
File::AptFetch::Simple (what has more) -- control progress
indication.
TEST-SUITE
Two things should be tested in regard with F::AF:
methods and structures
I wouldn't say I'm happy with this part (t/void/*.t). It's mostly
a crash-type testing. And a problem here is that I'm not
completely sure what an input (from methods side) here could be.
APT methods itself
Only local methods are tested. And *cdrom* method, among them, is
no way touched. Remote methods are missing from the test-suite.
This problem is easily fixable -- some kind Perlist could upload
on CPAN some modules, in Test namespace, what would provide fake
servers of http://, ftp://, rsh:// etc protocols. Anyone?
And one note. F::AF is absolutely useles outside APT based world (I'm
not about *.deb, I'm about APT). Then t/file/*.t and t/copy/*.t skip
completely. Meanwhile t/void/*.t enables some cute workaround, that
would provide all needed "apt-config(1)" functionality without that
executable itself. (Just in case on that platform there will be
Debian some day.) But we still stay within POSIX world -- no future
outside.
AVAILABILITY
pure distribution
and debianization stuff too
# TODO
subversion (root)
# TODO
BUGS
please report here
please don't report there
# TODO
COPYRIGHT AND LICENSING
* Copyright 2009, 2010, 2014 by Eric Pozharski
* AS-IS, NO-WARRANTY, HOPE-TO-BE-USEFUL
* GNU Lesser General Public License v3