At 09:39 AM 12/12/2001 -0500, Greg Hudson wrote:
>This might be subject to interpretation, but my reading is that there is
>some kind of Subversion-wide global default line-ending style, which can
>be overridden by a property, but we apply the global default style to
>the initial commit if the property is not set.
I hope that is not what was meant, since it mucks with the data much more
than would be necessary otherwise. But I guess we'll have to wait for
clarification on that.
>I didn't say I committed the file with no contents, just with no
>vectors.
Sorry for misunderstanding. That is good, though, since it simplifies the
problem.
> Presumably there would be some boilerplate as part of the
>vector graphic program's binary format (just like svndiff has a little
>ID string at the front). Moreover, I didn't say there were no line
>endings in the file, just that there was nothing resembling *a line
>ending which needed translation*. Perhaps the file happens to contain
>bare LFs but no CRs or CRLFs which would need translation to LFs.
Perfect. So when you checkout the file on Windows, the LF is changed to
CRLF and you have a corrupt file. The corruption might correct itself with
Ben's method (a faint hope) but it will not with your method.
>In step 2, when I commit the file with lots of contents, it is likely to
>contain, just by chance, CRs and CRLFs as well as bare LFs, due to the
>addition of lots of vectors.
Yes, but if this is the same platform you did the initial commit on, the
text-base would have matched the working copy before the edits. No
conversion on checkout (or update) should mean no conversion on commit,
don't you think?
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Oct 21 14:36:52 2006