[Bioperl-l] First cut svn repository [was Re: SVN and ...Re: Perltidy]

[Chris Fields]

On Jun 26, 2007, at 3:21 PM, George Hartzell wrote:

> ...
> If you have a dev.open-bio.org account and you're in the bioperl
> group, you're good to get at it via:
>
>   file:///home/hartzell/bioperl
>
> or
>
>   svn+ssh://dev.open-bio.org/home/hartzell/bioperl

I managed to get it working using file://.  Haven't tried svn+ssh yet  
but I've had persistent problems getting ssh to work properly on my  
macbook; not sure why yet but I haven't had time to play around with it.

> There are a couple of things to think about:
>
>   - how are we going to provide access.  I *think* that I heard a
>     decision to use http:// and https://.  Who gets to set that up?

That hasn't been decided yet and will be up to a consensus of the  
core devs, but I think the odds are in favor of allowing https:// but  
against allowing http://.

As for setup that could be anyone with admin privs, though it may be  
best left up to Chris D, Jason, or Mauricio.

>   - what do we want to do about keywords.  The cvs2svn tool guesses
>     and automatically sets the svn:keywords property to Author Date
>     Revision and Id on many of the files in the tree.  If it looks
>     like it got it right, we can stick with it.  Or, we can disable
>     that conversion and I've cribbed a little script that'll grep out
>     files using Id and set the svn:keywords property accordingly.

Probably again a consensus issue, but you can choose one route.  My  
inclination is the former if it's easier.

>   - what do we want to do about svn:ignore?  I haven't seen any
>     .cvsignore files.

Not sure.  I've never used one personally, but (as Jason suggests) if  
you have ideas for one you can propose them, or we can suggest devs  
set up svn::ignore locally.

> Beyond that, how does the repo look?

Seems fine, though a simple 'svn file:///home/hartzell/bioperl'  
checkout gets everything (all distros, branches, etc).  We need to  
make sure everyone uses 'svn co file:///home/hartzell/bioperl/bioperl- 
live/trunk /live' or similar if they just want the latest core/db/etc.

We'll also need to start a svn wiki page to show how to get relevant  
distros (similar in style probably to the cvs page, with dev  
information, how to set up ssh keys, https stuff, etc).

> How are we going to cut over?
>
> Are we going to try to push svn commits to the read-mostly CVS repo,
> or just keep it around for history's sake (I lean towards the latter).

I think a clean cut-over.  Everyone would be warned to hold commits  
for a day (lest they be lost), then probably do something in this order:

- switch cvs to read-only except for svn commits
- run a clean cvs2svn
- set up svn as read/write
- set up test commits to cvs via svn
- disable cvs commit messages to bioperl-guts, enable svn commit  
messages in it's place.
- push svn commits over to read-only cvs

cvs >>must<< be read-only after that point (no cvs->svn commits),  
with write access only available through svn.  If at some future  
point there is no reason to keep it around or that it is more trouble  
than it's worth, we can make a decision then on cvs's fate.

> g.

chris