Migration of Gentoo Repositories from CVS to Subversion.
========================================================

* Interested Parties:
	mail to: klieber,
	CC to: pylon, robbat2, trapni (that is me)
	(and maybe: absinthe - as he was doing the study AFAIK)

* Prerequirements:
	* make sure, client side and server side repositories
	  does not require to much disc space.
	* make sure performance is really better than (or equal to) CVS.
	* create an ebuild dev-utils/cvs2svn as it isn't done yet.
		DL URL: http://cvs2svn.tigris.org/files/documents/1462/16792/cvs2svn-1.1.0.tar.gz
		DEPEND="app-text/rcs"

* Convert the Repository:
	$ emerge -pv cvs2svn  # will install rcs if not yet done
	$ emerge -v cvs2svn
	$ mkdir /var/db/svn
	$ cd /var/cvsroot
	$ cvs2svn -s /var/db/svn/gentoo-x86 gentoo-x86

	in case cvs2svn aborts because of errors in pass 1, fix them by hand
	and resume at pass 2

	$ cvs2svn -p2 -s /var/db/svn/gentoo-x86 gentoo-x86

	Such an error maybe is:
		"ERROR: A CVS repository cannot contain both 
		 repos/foo,v and repos/Attic/foo,v"
	In this case, just remove the "repos/Attic/foo,v" or merge by hand.
	This error occurs ONLY when cvs admin had done some hand editing
	on cvsroot (I got told so on IRC)

	The whole conversion process took me up to 20 hours on
	my AMD AthlonXP 2800+ with 1GB RAM;

	*** Results ************************
	Well, okay, while the CVS tree just needs 1GB, the SVN tree even 
	needs 2.9GB with fs type bdb. So, lets see how fs type fsfs goes then.
	This bdb -> fsfs "fs-type" conversion may took me some hours, too.
	...

	*** server side disk usage *********
	du -ckhs /var/db/svn/*
	2.9G    portage-gentoo-x86-bdb
	2.0G    portage-gentoo-x86-fsfs

	*** svn checkout timings ***********

	svn co svn+ssh://trapni@subversion.trapni-akane.org/var/db/svn/portage-gentoo-x86-fsfs/trunk
	real    53m7.635s
	user    5m57.373s
	sys     2m10.571s

	cvs co gentoo-x86
	real    8m47.846s
	user    0m31.121s
	sys     1m1.356s

	see testing svn repositories:
		* http://subversion.trapni-akane.org/cgi-bin/viewcvs.cgi/?root=portage-gentoo-x86-bdb
		  with BDB fs backend type (complete conversion)
		* http://subversion.trapni-akane.org/cgi-bin/viewcvs.cgi/?root=portage-gentoo-x86-fsfs
		  with FSFS fs backend type (complete conversion)

* Anonymous Repository Access:
	* This is possible via Apache2 modules mod_dav_svn and mod_umask.
	* svn:// URL scheme supports anon access, too. 
	  But it's questionable wether it's required though and reslly 
	  a good advice to provide this, too.

* Developer Repository Write Access:
	* either via svn+ssh://
	* via http:// and https:// using following apache2 modules:
		* mod_dav_svn
		* mod_umask - to allow svn+ssh:// as well
		* mod_auth_pam - to auth against PAM accounts
			(and maybe install a pre-commit hook
			 in order to allow ACL based commits)

* post changes of 3rd level tools:
	* repoman: must understand local svn repositories
	* IRC Bot (CIA-10): needs to be notified via svn post-commit 
		hook on new commits;
		CIA-5 on #cvs2svn does obviousely do this for svn, so there 
		must already be a solution for this
	* Howto's needs talk about svn instead of cvs then, too ;)
	* create backup scripts (using svn hotcopy)
	* ...

* IRC NOTES:
	15:46 < trapni> btw, there's no option to like --fs-type in cvs2svn to pass :( 
	15:47 <@bliss> trapni: trunk has it. for older versions create it using svnadmin and pass --existing-repos to cvs2svn

* online references good to know about:
	* http://www.erenkrantz.com/oscon/OSCON%202003%20Apache%20HTTP%20Authentication.pdf
	  (apache authentification)
	* http://sam.zoy.org/writings/programming/svn2cvs.html
	  (contains good script excerpts on how to set svn:ignore and svn:keywords on a whole repository at once)
	* https://moin.conectiva.com.br/RepositorySystem
	* http://subversion.tigris.org/propaganda.htm

==== TIMINGS ====
$ cvs2svn -s svnroot/gentoo-projects cvsroot/gentoo-projects
....
pass 1:    32 seconds
pass 2:     0 seconds
pass 3:     0 seconds
pass 4:     1 second
pass 5:     2 seconds
pass 6:     0 seconds
pass 7:     0 seconds
pass 8:   187 seconds
total:    224 seconds

$ cvs2svn -s svnroot/gentoo-src cvsroot/gentoo-src
....
pass 1:    29 seconds
pass 2:     2 seconds
pass 3:     0 seconds
pass 4:     6 seconds
pass 5:    15 seconds
pass 6:     0 seconds
pass 7:     0 seconds
pass 8:  1329 seconds
total:   1383 seconds

Though, the following modules migrated w/o any problems

	gentoo-projects, gentoo-src

==== gentoo-x86 conversion errors =====

Error summary:
ERROR: A CVS repository cannot contain both cvsroot/gentoo-x86/gnome-extra/users-guide/users-guide-1.2.ebuild,v and cvsroot/gentoo-x86/gnome-extra/users-guide/Attic/users-guide-1.2.ebuild,v
ERROR: A CVS repository cannot contain both cvsroot/gentoo-x86/gnome-extra/users-guide/files/digest-users-guide-1.2,v and cvsroot/gentoo-x86/gnome-extra/users-guide/files/Attic/digest-users-guide-1.2,v
ERROR: A CVS repository cannot contain both cvsroot/gentoo-x86/dev-lang/entity/entity-0.7.2.ebuild,v and cvsroot/gentoo-x86/dev-lang/entity/Attic/entity-0.7.2.ebuild,v
ERROR: A CVS repository cannot contain both cvsroot/gentoo-x86/dev-lang/entity/files/digest-entity-0.7.2,v and cvsroot/gentoo-x86/dev-lang/entity/files/Attic/digest-entity-0.7.2,v
ERROR: A CVS repository cannot contain both cvsroot/gentoo-x86/media-sound/mpg321/files/mpg321-0.1.5-ao.diff,v and cvsroot/gentoo-x86/media-sound/mpg321/files/Attic/mpg321-0.1.5-ao.diff,v
Exited due to fatal error(s).

=== FIXING ===

cvsroot="/var/db/cvs"
pre_removes=(
	"gentoo-x86/gnome-extra/users-guide/users-guide-1.2.ebuild,v"
	"gentoo-x86/gnome-extra/users-guide/files/digest-users-guide-1.2,v"
	"gentoo-x86/dev-lang/entity/entity-0.7.2.ebuild,v"
	"gentoo-x86/dev-lang/entity/files/digest-entity-0.7.2,v"
	"gentoo-x86/media-sound/mpg321/files/mpg321-0.1.5-ao.diff,v"
)
for file in ${pre_removes[*]}; do
	rm -f $cvsroot/$file
done

vim:ai:noet:ts=4:nowrap