|
=== Release 2.3.3, October 28, 2002
Changes to start coping with MKS Source Integrity conversions.
This change allows genmetadata to simply ignore (and hence, to
continue working!) the "ext" and "format" keywords that MKS Source
Integrity apparently adds to the rcs repertiore. This could possibly
be recoded to deal with arbitrary new keywords, but I don't want to go
there now.
Note that there *may* be other gotchas when attempting to use cvs2p4
to convert MKS SI repositories, which may be addressed by future
changes to cvs2p4:
1. It appears that MKS may compress (gzip) binary ,v files, but *not*
rename them to refelct this. Very odd.
2. It appears that MKS may allow a larger character set (than pure RCS)
in legal symbolic names. I can probably fix cvs2p4 to deal with these
pretty easily, but we should also consider whether any of the MKS names
might be illegal in Perforce (and thus needs some sort of
translation).
Intrepid users setting out to use this tool to convert MKS into
Perforce are encouraged to send problem reports to
"opensource@perforce.com".
|
|
=== Release 2.1, July 10, 2002
Change to handle RCS branch tags (so this tools can work with
RCS (vs CVS) repositories, too!) RCS branch tags are those
with an even number of "."s.
=== Release 2.2, July 18, 2002
You can now choose the Perforce filetype you'd like to have ALL of
your text files imported with. As long as it's "text" or "ktext" :-).
See the "$TEXTTYPE" configuration item.
For "developers": Added a new test/getgood script to grab the test
comparision files which have changed with a "runtest -gengood". Less
tedium.
Made the test/phone.gif,v a binary (-kb) RCS file, so we really test
that.
Fixed a bug in dochanges, wherupin the command to create a symlink to
the RCS archives was being created in copy mode, provided the copy did
not fail. Reported by John Wong (jw at unicom.com.au).
Added a new config item "DEADSTATE", for saying what RCS state value
means "deleted" in Perforce. It defaults correctly for CVS, so if
you're converting real CVS, you can just leave this as-it.
=== Release 2.3.1, July 30, 2002
Now always run p4d with an explicit p4d -r, in case the user's
environment has a P4ROOT set.
genmetadata now reports both the CVS file archive path and the
converted file path as it processes each file. This makes it easier to
tell what RCS archive file caused a problem if the file cannot be
parsed properly.
The scripts now use the more modern tie()/untie() database interface,
instead of the older dbmopen()/dbmclose() interface. The scripts
assume the presence of the DB_File (Berkeley DB 1.x) module. This
means that you are no longer left to the whim of the dbm support
compiled into your perl. Certain dbm implementations, notable that in
recent Solaris, caused problems. If you don't have DB_File support for
your perl, you'll need to find the tie() statements in the bin/*
scripts, and modify them to use some dbm package that you do have on
your system.
$P4_DEPOT can now be set to something that doesn't start with
"//depot/", and things will work properly (a new depot specification
will be created in the converted database).
Added README notes about conversions that merge into an existing
depot.
=== Release 2.3.2, September 3, 2002
Add $DEPOT config item, allowing RCS archives to appear as
$P4ROOT/$DEPOT/$CVS_MODULE; I.e., with $DEPOT set to "depot/IMPORT",
the RCS archive files will end up under
"depot/IMPORT/<modulename>/<filepath>"; With $DEPOT set to "depot",
they will be, simply, "depot/<modulename>/<filepath>". This makes it
easier to make later merges of subsequently imported changes easier.
Changed the database open code to allow setting the size of hash
buckets (allows large conversions without running out of space for
hash table pages).
Handle label names with '#' or '@' by substituting "_hash_" and/or
"_atsym_" into the converted label name.
(Thanks to Marc Tooley of Perforce Support for supplying these
imporvements).
|
