> From: Greg Hudson [mailto:ghudson@MIT.EDU]
> Sent: Tuesday, March 11, 2003 9:23 PM
> (I wish there were a mail archive of this list which kept message IDs,
> so I could generate proper responses after accidentally deleting the
> mail I want to respond to.)
>
> Branko wrote:
> > Whichever code it is, it's doing the wrong thing. The text base
> > should be converted to WC style, not the other way around, otherwise
> > you won't notice newline-only changes.
>
> > Oh, duh -- that would make keyword expansions cause problems for us
> > again. What a bore.
>
> This just means we have to parameterize the keyword expansion code so
> that we can say "expand this text base with the same substitutions as
> we used on the working copy."
We can simply translate the working copy to collapsed keywords. That
is already part of the API.
>> Hmm, maybe the conversion from WC style to text-base style should
>> notice if the WC file line endings aren't consistent with what's
>> declared in svn:eol-style?
>
> I think it used to, but we turned that check off for everything but
> "svn commit" after someone complained that it was difficult to recover
> from having inconsistent newlines. You couldn't even "svn revert" at
> that time, which was really bad, but you also couldn't "svn diff" to
> find out where your problem was. Of course, it sounds like now, you
> can successfully "svn diff", but it won't tell you where your problem
> is.
Right, so we need to do the translations the other way around and
compare in working copy style. Anyone care to make that change?
Or file an issue?
Sander
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Mar 11 21:29:09 2003