Maybe it makes sense to add a --case-sensitive-filenames switch
to the client under win32 that means that the exact filename casing
should be used on commit regardless of whether another file with
similar casing already exists. Otherwise, the commit should apply to
the preexisting file with different casing. This is the expected
behavior coming from a win32 paradigm. Also, I don't think you should
handle casing issues with a warning and then present the user with a
choice; I think it should result in an error unless the appropriate
switch is used (i.e. --case-sensitive-filenames).
Gili
On Tue, 21 Dec 2004 14:34:19 -0800, Peter Yamamoto wrote:
>Actually it's really a matter of interpretation whether or not the
>different case should represent a "conflict" or not. To be completely
>flexible, you'd expect the platform dependent client to issue a warning
>and a chance for the user to choose which interpretation (and possibly
>local renaming to resolve the ambiguity).
>
>The fact is that the Windows client may potentially not be able to
>represent the repository when case sensitive filenames are used... which
>is a good reason to be able to turn this "feature" off on the server
>side.
>
>Peter
>
>-----Original Message-----
>From: Peter Yamamoto [mailto:peter.yamamoto@page44.com]
>Sent: Tuesday, December 21, 2004 2:22 PM
>To: Gili; Norbert Unterberg; Peter N. Lundblad
>Cc: users@subversion.tigris.org
>Subject: RE: Re: Renaming files on win32
>
>It's not just a unix vs win issue either... we had a similar problem in
>a completely win environment where the application software supported
>crossplatform clients and hence was case sensitive.
>
>I can't recall the exact circumstances and tools, but we had case and
>'duplicate' file/folder name issues when using StarTeam which is a
>cross-platform version control package.
>
>For example, one person would create/add a file in a folder (for example
>a texture file generated at some point by some artist). Later another
>artist may work on the file and whether through tools or simply because
>of having to retype a name, use a different casing. Ultimately a file
>with the same (intended) contents but different casing is on the drive.
>
>According to the server (which supports mixed case) this file is
>different than the file already in that folder and hence can be *added*.
>Eg they never checked out the original version, and this "new" file
>technically was different from the same file from the server's point of
>view.
>
>On the server side all is "fine" so to speak. But back on the client
>side, the file would continuously be in "modified/conflict" since when
>it got checked out, the software would say it's a different version than
>the other named file [since at this point the local workspace
>technically does have "both" versions being represented in the same
>file!-P]. Of course the real problem should have been that it should
>have shown up as in conflict in the first place before allowing the add.
>
>I guess the point is that you have to be very careful about the
>layers/operations where you try to shield windows clients from this
>anomaly, while still trying to preserve the capabilities of the server
>side representation of the repository.
>
>Peter
>
>-----Original Message-----
>From: Gili [mailto:junk@bbs.darktech.org]
>Sent: Tuesday, December 21, 2004 2:04 PM
>To: Norbert Unterberg; Peter N. Lundblad
>Cc: users@subversion.tigris.org
>Subject: Re: Renaming files on win32
>
>On Tue, 21 Dec 2004 22:57:42 +0100, Norbert Unterberg wrote:
>
>>I even do not see the need to drop case-sensitivity in subversion
>>(server part) to make the windows svn client behave as a windows user
>>would expect (at least for the working copy). Since svn stores a copy
>>of the file with the original file name in the text-base, the client
>>can detect the original name (as stored in the repository) and identify
>
>>the file, then take the action required to recognize the file, move,
>>checkin, rename etc, even if the case of the name changed. This should
>>be possible without server changes, shouldn't it?
>>
>>Norbert
>>(not knwoing too much details about svn internals)
>
> Yes, but I am concerned about what happens when someone checks
>in different files "abc" and "Abc" with a Unix client and then someone
>under win32 tries to check them out.
>
>Gili
>
>
>---------------------------------------------------------------------
>To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
>For additional commands, e-mail: users-help@subversion.tigris.org
>
>
>---------------------------------------------------------------------
>To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
>For additional commands, e-mail: users-help@subversion.tigris.org
>
>
>---------------------------------------------------------------------
>To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
>For additional commands, e-mail: users-help@subversion.tigris.org
>
>
Received on Tue Dec 21 23:51:39 2004