kfogel@collab.net writes:
> Philip Martin <philip@codematters.co.uk> writes:
>> We have client side code that prevents those keywords being set on
>> binary files. As far as I know, it doesn't prevent someone setting
>> the keywords on a non-binary file, and then changing the file to
>> binary. It's not clear to me whether we should ignore the keywords on
>> a binary file. If we do choose to ignore them I am a little uneasy
>> about the behaviour being distributed in lots of places in the code,
>> far better to pass some sort of "binaryness" indicator to
>> copy_and_translate.
>
> I don't like the idea of ignoring them at all. Surely there are
> binary formats in the world that can tolerate (say) keyword expansion.
I suspect such formats do exist.
> This does not contradict the client-side user protections. Having
> client-side code to prevent setting translation properties on binary
> files makes sense, because it's far more likely to happen accidentally
> than not. One can still override with -F, or by a workaround of
> temporarily resetting the mime-type, if they really need to.
>
> Which means that if we *do* receive a translation property on a
> non-text file, someone probably went through extra trouble to put it
> there! Therefore, we shouldn't ignore it.
As far as I can see update will apply translations to binary files if
the appropriate properties are set. That means it's probably a bug
that svn cat behaves differently.
--
Philip Martin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Nov 17 00:09:08 2003