Hello Karl,
Thanks for the prompt response.
Karl Fogel wrote:
> Changing data on commit from a hook is highly discouraged, because it
> means that what's in the repository will be different from what the
> client thinks it committed...
>
>
Are there any cons besides the working copy inconsistency? I pretty much
don't care about this thing as the property is only changed by the
script and fetched when required from the repos directly rather than
from a working copy.
But I'd love Subversion to handle such situations better than it
currently does ;)
> I'm afraid I don't have an answer for your question, and could only
> recommend the usual debugging techniques. One problem might be that
> when you copy a directory in Subversion, it doesn't actually make
> copies of the things below that directory -- they haven't changed, so
> they're just included by reference. Read about the "bubble-up" method
> for more on this.
>
> http://subversion.tigris.org/design.html#server.fs.struct.bubble-up
> http://subversion.tigris.org/hacking.html#docs
> http://www.red-bean.com/kfogel/beautiful-code/bc-chapter-02.html
>
>
Thanks for pointing out to the interesting reading. I'm not sure I'm
proficient enough in hacking to fulfill such a bug hunt :) Nevertheless,
I'll take a shot.
BTW, this may be interesting: I've discovered that a Berkley DB-based
repos doesn't show this discrepancy. So, it seems that in the fsfs
plug-in where something unexpected happens.
Regards,
Dmitry
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Thu Oct 11 08:48:35 2007