Re: [Issue 2746] New - update overwrites w/o warning local modification if local timestamp did not change

Malcolm Rowe <malcolm-svn-dev@farside.org.uk> writes:
> Er, but wouldn't that be the common case - when updating an unmodified
> file, for example?

Oh, you're right. That *is* the common case. Urgk. Right, now I
remember that we went through this debate when we originally decided
to rely on timestamps.

I may be about to redefine the "sacred" in "data is sacred". I mean,
there are limits! :-)

> Currently (I'm guessing) we read the text-base and apply the delta to
> produce a new text-base, then copy it over the existing wc file and
> rename it over the text-base. Can we instead read the (believed
> unmodified) wc file and apply the delta to that? We can already check
> the MD5 of a file while we're reading it, so that wouldn't increase the
> amount of data we need to read.

Normally, we don't necessarily read all the bytes in a file when we
use it as the base for a delta application, though.

> (And in the case where we believe the file is modified, we'd continue to
> use the pristine text-base).
>
> Oh, wait, that won't work for translated files... and I'm not sure
> whether we even hold an MD5 of the translated version of the file, so I
> don't know how we'd solve this problem for those files anyway.

If we don't hold a checksum of the working file, we could -- I don't
see why that would be hard. But (see above) it wouldn't help us much,
at least the way delta application currently works.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Mar 19 18:34:06 2007