[Bf-committers] SVN restructure

jonathan ferguson jdpf at edumetrics.org
Fri Jun 1 18:18:56 CEST 2007


hi

On May 29, 2007, at 6:25 AM, Nathan Letwory wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> A workflow for a proper checkout would thus be:
>
>> mkdir mycodingdir
>> cd mycodingdir
>> svn co URL/trunk (or svn co URL/trunk/blender blender)
>> svn co URL/lib/windows lib/windows

Although I don't really have a better solution, I comment that this  
has the unfortunate side-effect of breaking the expected workflow of

cd bf-blender/trunk/
svn update

and having *all* of the relevant files including the lib tree,  
updated. Improperly updated lib trees can cost quite a bit of time  
when building blender suddenly breaks (and you have no idea why  
because code that depended on libxyz0.1 has changed to depend on  
libxyz0.2 or some such thing and you haven't been reading every Bf- 
blender-svn message).

Then the workflow for "day-to-day" updates is really:

cd bf-blender/trunk/
svn update
cd bf-blender/lib
svn update

Trying to isolate by hand bf-blender/trunk/blender and bf-blender/ 
trunk/lib/<target>/ as separate check-outs also breaks the idealized  
workflow of subversion. Trying to svn update in such a "hand-built"  
directory tree will cause svn to puke out that "bf-blender/trunk/lib  
is unversioned" or some such. For this reason, I gave up trying to  
isolate the lib tree, and just checked out the whole thing. :| I  
don't know of a way to tell subversion to mask a checkout. Anyone know?

Idealized, the workflow should really just be:

cd bf-blender/trunk/
svn update

As I say, I haven't a better solution for helping people isolate the  
lib tree on checkout, but wanted to articulate the consequences of  
jesterKing's suggested change here. Other experiences?

have a day.yad
jdpf






More information about the Bf-committers mailing list