META.yml #32 

# http://module-build.sourceforge.net/META-spec.html
#XXXXXXX This is a prototype!!!  It will change in the future!!! XXXXX#
name:         P4
version:      3.5869
version_from: P4.pm
installdirs:  site
requires:

distribution_type: module
generated_by: ExtUtils::MakeMaker version 6.17
# Change User Description Committed
#32 5869 Tony Smith Fix memory leaks.
Thanks to John LoVerso and Sandy Currier for
tracking these down and sending me the patch.
#31 5868 Tony Smith Port P4Perl to Perl 5.8.8.
This change is spectacularly ugly,
but then so are the innards of Perl. See the long thread at:

http://www.nntp.perl.org/group/perl.perl5.porters/2006/06/msg114383.html

for details of the problem, and some discussion of solutions. I've
had to come up with a solution that doesn't involve patching people's
Perl installations, so my fix is even less easy on the eye but it
appears to work, and hopefully hasn't broken things for older
Perl versions.
#30 5787 Tony Smith Update P4Perl to build against the 2006.2 API, and detect
(and reject) the passing of tainted data to P4::Run() as
a security measure.
#29 5708 Tony Smith Add static P4::Identify() method to report the version of
P4Perl, and the API used to build it.
#28 5704 Tony Smith Fix bug introduced in change 5677: the quoting of the const_char
definition wasn't good on Windows as it doesn't really understand
single quotes. Switched to using double-quotes.
#27 5692 Tony Smith Add support for $p4->ServerLevel() which returns the server's
'server2' protocol level. This is not the same as, but is
closely aligned to, the server version and can be used
to test for feature availability. If you need explicit Perforce
version strings, run 'p4 info' in tagged mode and parse the
'serverVersion' string.

New feature requested by Robert Cowham.
#26 5677 Tony Smith Rework P4Perl build script to support 2006.1 API.
There were
some sweeping changes in the 2006.1 API which did away with the
old const_char definition. Unfortunately, since P4Perl has to
build with older APIs, I can't quite do the same. This change
tweaks the way that Makefile.PL determines the definition for
const_char so that (a) it's always 'const char' when a 2006.1
or later API is used, (b) it uses the platform specific hint
for older APIs (c) it falls back on 'char' in the absence of a
hint.

In an ideal world, we'd be able to compute the options based on
(a) OS, (b) compiler version (c) Perl version and (d) API version,
but that's tough to get right.

No functional change.
#25 5624 Tony Smith Add hints file for Mac OS X from Tim Bunce (thanks Tim!).
#24 5592 Tony Smith Bug fix: We were erroneously making a Perl scalar mortal (which causes
its reference count to be decremented when it goes out of scope) when
receiving binary data from the server. Decrementing the reference count
to early meant the scalar was being cleaned up when it was in fact
still in use.

This change simply removes the sv_2mortal() call.

Thanks to Mike Hall at National Instruments for finding this one.
#23 5564 Tony Smith Fix memory leaks when dealing with tagged output with array
members. One of these days I'll get the hang of Perl's pesky
reference counting mechanism...

These leaks would have been most noticable with 'p4 filelog' (and
that's where they were reported), but in fact any command that
returned tagged output with array members would have leaked.
#22 5396 Tony Smith x86_64 porting changes.
Use INT2PTR and PTR2INT to handle the
stashing of the PerlClientApi pointer in a Perl scalar. This is
necessary because, despite all documentation to the contrary, an
I32 is not 64-bit capable on all 64-bit machines.

Also, the hints file now looks for x86_64 in the architecture
name and if it finds it, then const_char='const char' rather
than the default of 'char'.
#21 5313 Tony Smith Add new SetApiLevel() method to allow users to lock scripts to a
particular API level. This helps when upgrading to new servers
that extend support for tagged output to hitherto unsupported
commands (2005.2 did a lot of that).

See the C/C++ API Release Notes for the full details, but by way of
example, to lock scripts to the 2005.1 interface use:

    $p4->SetApiLevel( 57 );
#20 5259 Tony Smith Update P4Perl for 2005.2 API changes.
The 2005.2 API supplies forms
ready-parsed to the client when used in tagged mode. This is fine for
P4Perl, except that we were not caching the specdef if no parsing was
required and that meant that although forms could be converted to
hashes, the reverse direction was broken.

This change makes sure that we cache the specdef whenever it's
available.
#19 5073 Tony Smith Yet another P4::SetProg fix.
Turns out the last change fixed
everything except the build script and it all worked...
unless you used the 2005.1 API which has a different format
for the Version file.

P4::SetProg is a no-op when P4Perl is built with an older API
so not identifying the API build properly breaks it thoroughly.

This change adapts P4Perl's setup script to support both
pre-2005.1 and 2005.1 formats, and gives the user the
chance to enter the API version manually if it can't be
automatically determined.

Installers also included in this change
#18 5067 Tony Smith Bug fix: P4::SetProg() interface method was missing so SetProg
wasn't working too well!
#17 5038 Tony Smith Bug fix: Fix memory leaks in P4Perl reported by Craig Galley.
Perl's
reference count garbage collection is not much fun to work with,
but hopefully this change plugs P4Perl's leaks. There's still a
leak that remains, but whether it's in P4Perl's code or just in
Perl I don't know. A loop like this:

    while( 1 ) {
    my $p4 = new P4;
    }

will leak like a sieve but I'm pretty sure P4Perl is cleaning up
as it should. While it's very difficult to be certain with Perl's
memory mode, creating one P4 object and using it multiple times
now appears to be pretty steady.

Also fixed use of uninitialized debug variable which could produce
debug output you hadn't asked for.
#16 5035 Tony Smith Bug fix: call ClientApi::SetProg() before every command instead of
just once as this value is not retained by the Perforce API.
#15 4987 Tony Smith Bug fix for tagged mode output from 'p4 diff2'.
Diff2 is one of
the few (only?) commands to use variables of the form 'var' and
'var2' rather than 'var1' and 'var2'. Normally, if there's no
numeric suffix to a variable, P4Perl can assume it's looking at
the only instance of that variable in the output. In the case of
'p4 diff2', that's not true.

This change enables P4Perl to adapt to this change of circumstances
and convert a previously scalar member of the hash result into an
array member.
#14 4873 Tony Smith Bug fix: fix typos in test harness that were causing the tests
to fail.
#13 4864 Tony Smith Bug fix: Introduce workaround for obscure 2000.1/2000.2 protocol
bug that I really thought we'd seen the last of.

Along the way, a total revamp of the debugging output - required
to diagnose the problem.
#12 4856 Tony Smith Rework P4::Errors() and P4::Warnings() so that they return a list
rather than an array. Perl seems to like this more and it's easy
to assign the list to an array should you wish to do so.

Note that this may cause some backwards-compatibility issues.
#11 4831 Tony Smith Change implementation of P4 class from being a blessed reference
to an integer (pointer) to a blessed reference to a hash. The
pointer is now stashed in a member of the hash. This makes it easier
for those that want to to subclass the P4 class and bolt on their own
functionality. No functional change.
#10 4804 Tony Smith Add support for P4::SetMaxResults() and P4::SetMaxScanRows() which
specify the desired limits for an instance of the P4 class. Note
that the limits remain in force until disabled by setting them to
zero.
#9 4754 Tony Smith Add support for passing multiple items of input to Perforce commands
that need them. The prime example is 'p4 password' which prompts the
user three times for password input (old password, new password and
new password again).

Also add a P4::Password( $old, $new ) method to make it nice and
easy to use.
#8 4698 Tony Smith Bug fix.
Correct client initialization so that it no longer causes
problems if the connection to the server fails for some reason.

Also corrected the number of tests in the test harness.
#7 4676 Tony Smith Enable P4Perl to work against a server in unicode mode.
This change
adds two new methods to the P4 class: SetCharset() and GetCharset()
which have the expected behaviour.

Thanks to  Raymond Danks <raymond.danks@amd.com>.

Also cleaned up the test harness a little.
#6 4667 Tony Smith Caught by the old "hadn't saved my changes" problem, so the
Changes file in the last submission was incomplete. Bumping
the version number to retry...
#5 4666 Tony Smith New ParseSpec() and FormatSpec() methods allow you to convert
specs between hash and string representations easily. Shortcut
methods Parse* and Format* are also defined. (i.e. FormatClient()
and ParseLabel() etc.)

New methods IsTagged() and IsParseForms() tell you if your client
is in tagged/form parsing mode respectively. If you care.

P4::Tag() is deprecated in favour of P4::Tagged(). P4::Tag() exists
for backwards compatibility
#4 4608 Tony Smith Bug fix: The SetInput() method was omitted in the big rewrite
so quite a lot was broken in builds 3.4579 and later. This
change fixes that omission, and adds support for 'p4 login'
too (that was how I discovered that SetInput() was missing).
#3 4586 Tony Smith Update P4Perl tarball with most recent change.
No need for a new
installer as there's no functional change.
#2 4583 Tony Smith Update distribution files to include Windows porting changes
#1 4579 Tony Smith Rewrite P4Perl to be more like P4Ruby.
This change does away
with the old P4/P4::Client split and pulls all the functionality
of P4::Client into P4. Hence P4::Client is now deprecated.

There are a few gotcha's - see the Changes file, and documentation
for the details, but in general it's backwards compatible.

As part of this change, I'm also releasing the previous current
 versions of P4 and P4::Client as released versions - for posterity.
P4 now gets a v3.x version number so the old versions will stand out
clearly.

Hopefully it's all working - works fine for me - but people should
consider this a beta build, for now at least.