On 11/24/05, Peter N. Lundblad <peter@famlundblad.se> wrote:
> On Thu, 24 Nov 2005 dionisos@tigris.org wrote:
>
> > Author: dionisos
> > Date: Thu Nov 24 12:41:02 2005
> > New Revision: 17514
> >
> > Modified:
> > trunk/subversion/libsvn_client/blame.c
> >
> > Log:
> > Fix issue #2431 (blame ignores svn:eol-style).
> >
> > Tested by: zhakov
> >
> > * subversion/libsvn_client/blame.c
> > (file_rev_baton): New members eol_style and eol_str for storing
> > the requested file translation type.
> > (get_eol_style): New. Function to adjust eol style from propdiff sent.
> > (file_rev_handler): See if the file needs translation for the eol style,
> > translate if it does.
>
> Have you tested if this has any performance impact for, say many revisions
> or large files? I'm concerned because blame fetches a lot of files and
> diffs them. If this is a real problem, can't we just add the correct EOL
> sequence when outputting the lines (after stripping whatever line endings
> there are on the actual lines.)
That depends on what you want to achieve?
> It seems strange to translate each and
> every revision of a file just to get the correct EOL characters in the
> last one. (Or isn't there a streamy interface to translation that could
> be used to avoid the extra tempfile?)
Heh. well, I've been thinking about that too. No, there's no stream to
achieve that, but if there were, we could have 2 new stream
interfaces: one splitting 1 stream into 2 streams (for output), the
other to translate output when written through the stream: that way we
could prevent a lot of io on update, merge or switch, or at least, it
looks like it at first sight.
I'd love to create the translating stream interface and made a note to
do so, but this is what I could come up with for now. Maybe we should
ask DannyB to do testing on the gcc ChangeLog file?
bye,
Erik.
Received on Thu Nov 24 21:56:20 2005