On 10.11.2011 16:21, Neels J Hofmeyr wrote:
On 11/07/2011 07:14 PM, Miha Vitorovic wrote:
On 7.11.2011 16:08, Neels J Hofmeyr wrote:
Can you argue up a case where one would want a non-revision-pegged external
excluded from commit? I'm reluctant to take simply previous externals
behavior as argument, because externals have always sucked so far.
I can :)

I spend my days writing "code" in LabVIEW. In short, it's a graphical
programming language. Its files are a sort of combination of source code and
binary. We have our projects organized around a common framework The
framework is included in the projects using externals. Don't ask me why, but recompiling a project also recompiles some framework files. As a result this marks them as modified for Subversion. And when committing the project we
really don't want to have those framework files committed as well.

Agreed -- now: would you be OK if I told you: to omit those dir externals
from commit, you have to either supply --exclude-externals, or you have to
put the external on a specific revision, say r123, like:

  svn ps svn:externals "^/framework-src@123 bin/fw" .

Like that you could choose for each external: do I fixate it to a specific revision, so that it is not committed automatically? Or do I leave out the
revision number, thus the external is included in commit recursion?

Whenever a newer framework should be used, you have to modify the revision number in the externals definition and commit that. Subsequent updates then
bring the newer framework externals into all colleagues' WCs.

Would you be fine with that?


Sorry for not replying sooner - my sentiments were simply not strong enough either way. But there is something to be said about changing default behavior. And I think others have done a splendid job doing that.

In short - whatever you (the developers) choose will be fine. I have a feeling default behavior will remain as it is, though :)

Br, Mike

Reply via email to