> From: Philip Martin
> Sent: Wednesday, February 26, 2003 4:20 PM
[...]
> I guess we will just have to fix the bugs.
*nod* Even though very nasty :(
> There appears to be way to get CRLF line endings into the repository
> together with an svn:eol-style of native. Try
>
> svn co -r5007 http://svn.collab.net/repos/svn/trunk/build/win32 5007
>
> and look at 5007/.svn/text-base/build_neon.bat.svn-base, it has CRLF
> line endings, but the svn:eol-style is set to native.
Ah, right, that's why Branko changed them from CRLF to LF in 5008.
[...]
> before being sent to the repository. From what Bill has told me
> that's how it happened in the past. The real problem is that this
> illustrates how fragile the system is, it relies on the "good
> behaviour" of the clients. Is it going to work when plain WebDAV
> clients can commit to a Subversion repository?
Urf, I guess not. Not unless we are going to implement guards
_server_ side preventing CRLF from being committed on 'native' style
files (same goes for other eol-styles).
> Having a file like this causes problems, the merge bug in this thread
> being one.
Uhuh
> When the next change to the file's text is committed, the
> client will convert all the CRLF to LF, so the commit will change
> every line even though the user may only have changed one.
Ouch!
> Running 'svn diff' on the working copy before the commit doesn't show the
> problem, because diff works on files in the working copy form, not in
> the repository form.
So either we switch all operations to always work on wc form or we switch
to always operating on repository form, so that the problem is at least
noticable.
Sander
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Wed Feb 26 16:48:47 2003