Andy Levy wrote:
> On Tue, Sep 2, 2008 at 20:49, Craig McQueen <mcqueen-c_at_edsrd1.yzk.co.jp> wrote:
>
>> Simon Large wrote:
>>
>> 2008/9/2 Nathan Smith <nsmith_at_galileoprocessing.com>:
>>
>>
>> I do have a related question though. I really like Tortoise's diff and
>> merge interfaces, but I am also quite comfortable using the command line
>> for anything we need. I'm working on installing svn on the webserver to
>> facilitate quicker day-to-day operations. Right now I'm using Tortoise
>> with mapped drives to smba shares. In the configuration, it remembers
>> the 'Y' drive, not a path on the server. I can see some issues coming up
>> with different path names.
>>
>> Here's the question: How can I use both the command line and the mapped
>> drives on the same working copy? Does anyone know a good workaround?
>>
>>
>> You can use TortoiseSVN and the *Windows* command line client on the
>> same working copy. You must not use the Linux command line client on
>> the same working copy as TortoiseSVN. The working copy is assumed to
>> be local and subversion does not guarantee compatibility between
>> different OSs. You will end up with a broken WC if you do this.
>>
>> Simon
>>
>>
>> If there is a working copy incompatibility between OSs, that's too bad. Is
>> it too difficult to make WCs platform independent?
>>
>
> The primary incompatibility is with EOL markers. Windows uses CRLF,
> most of the rest of the universe uses LF. So if you have files marked
> with svn:eol-style=native, you may have trouble working with files in
> Windows which were last updated/checked out from the Linux side.
>
> There are also possible issues with file permissions, at the filesystem level.
>
>
>> I suggest that at minimum, SVN clients are able to detect that a WC is
>> incompatible, and report a sensible error message, rather than giving
>> obscure behaviour and errors.
>>
>
> The WC format itself isn't incompatible (that I've heard), it's
> differences in how the files are handled. And Subversion can't
> reliably detect that & protect you from yourself.
>
>
>
I don't believe "it can't be done".
If the WC were to store the EOL marker in use on the system, then it
could be checked by the SVN client every time it runs. If there's a
discrepancy, then report an error message. Something similar could be
done for permissions. Or keep it simple -- record the OS flavour in use,
maybe as simple as Windows vs *nix.
I know -- most of us think that a workflow that uses a WC on multiple
OSes is "weird". Still, it's amazing what schemes us human beings can
dream up and think "it should work, hey?" and then post a message to a
certain forum asking why it doesn't work. :-)
Regards,
Craig McQueen
Received on 2008-09-05 03:15:55 CEST