Should I Change To Bazaar

[A. S. Budden]

2009/1/16 Adrian Wilkins <adrian.wilkins at gmail.com>:
>> On most (but not all) work projects, I
>> use PVCS for the collaborative part
>
> You should be fine to use Bazaar in parallel with PVCS as long as you can
> get it to ignore .bzr - only one at the top level, not one for each folder
> like svn. Since it ignores .svn I presume it can ignore .bzr .

PVCS ignores everything unless you explicitly ask it to include
something.  If you do an "add folder", it adds everything, with no
ignore mechanism at all.

>
>> I use Windows and Linux (work and home) and (at the moment) the same
>> repository for both.  For the vast majority of projects I work on, it
>> is fine for LF line-feeds to be used throughout: the Windows tools
>> cope, so I use cygwin subversion as it keeps things nice, easy and
>> consistent.
>>
>
> Bazaar, like Subversion, doesn't mess with line-endings. There's a lot of
> discussion around line-ending support but for me I always prefer to just use
> Windows tools that support LF endings.

Excellent!

>> I run svnserve and
>> connect to the repository with svn://localhost.  This simplifies
>> things as I often log onto different Windows computers with the
>> working folder on the network, but the Toughdrive mapped as a
>> different drive letter:
>
> For windows I use the excellent USBDLM service which keeps a consistent
> drive letter for my thumbdrive.  [1], but bzr serve would also achieve the
> same effect. My chosen GNU/Linux distro (Ubuntu) uses consistent mount names
> when automounting USB media.

I generally find that windows uses consistent mount points for drives
(particularly after I've assigned one with Windows Disk Manager or
whatever it's called), but my problem is partly laziness: I tend to
use a _lot_ of different computers and I'd rather not have to either
use Windows Disk Manager or install (and pay for in the case of a
business licence) USBDLM on every computer.  svnserve saves me the
effort.  Anyway, bzrserve sounds very good to me.

>
>> shortens the URLs: the one thing I miss from CVS
>> was bash tab completion when typing "cvs co".
>>
>
> Because I use the filesystem direct, I get tab completion in any of the
> windows shells I use. You don't need to explicitly use "file://" URLs for
> Bazaar, it will infer them from paths.

I take it that (like svnserve) there's no bzrserve support for this?

>> The trunk contains all of the generic
>> code (drivers for the peripherals on the development board etc) and
>> some outline "App" code for the specific stuff.  I currently do "svn
>
>>
>> * Easier merging of core files across trunk and branches - keep
>> specific files in sync while changing others independently
>>
>
> You'll have to do this either by cherry-picking merges (not conducive to
> automatic merging) or by maintaining separate branches for "core" and
> "non-core" files. DVCS systems place more emphasis on being "versioned
> atomic trees" rather than SVNs "atomic changesets on a versioned
> filesystem".
>
> I would move the project-specific files out into their own branches. The
> whole "folder full-o-branches" concept needs some love though - there are
> some plugins to manage build groups but they are relatively young.

Okay, thanks for the advice.

>> * Commits should go directly to the repository on the Toughdrive (as I
>> think of the commit partly as a backup as well as a version-management
>> system)
>
> I tend to use "bound" heavyweight checkouts in this role. The revision
> history is kept locally in repository local to the working tree (in the .bzr
> folder of the tree, or a parent folder with a shared repo), and also
> committed to the remote repo (in this case, a USB drive). If you lose your
> USB drive (or just leave it at home that day), all the revision data remains
> intact at the client machine, you can continue to make commits and sync
> later.

Sounds very promising.

>> * Two parallel central repositories.
>>
>
> It's simple to push branches from one repository to another because of the
> underlying design ; just set a job to do it ; in the crontab or perhaps a
> job that occurs when the drive is mounted. Maintain two separate branches
> ("home" and "nothome") for your .vim folder and merge changes between them
> (you need a working copy to do merges).

I think I need to think about that one for a while...

> .... The above is actually "how it was". Since then, my organization has
> elected to prohibit writing files to unencrypted USB media, and enforced
> this with compulsory software. Now I use my laptop as my thumbdrive and run
> `bzr serve` when I need access to my branches from another machine. On my
> laptop I use lightweight (no repo) checkouts of my central repo ; on other
> machines I use heavy ones (with-repo). This continues the "distributed
> backup" theme and distributes my revisions to all the machines I access them
> on.

Makes sense.

Thanks for all the advice, I'll do a bit more reading and may well
bite the bullet once I've got my Network Drive Pro.

Al