On Mon, 2006-06-02 at 23:21 +0100, Martin v. Löwis wrote:
> Sure I would implement it. I would need to find the proper layer of
> abstraction, and then hope that I don't need to change much code at
> all. It must be somewhere in the working-copy code.
>
> As for what the benefits are: if done well, it works on all systems,
> not just Win32, and on all file systems, not just NTFS. Furthermore,
> it doesn't require hard-coding of file system conventions into the
> subversion code, which apparently is an important property in
> maintaining subversion. Finally, there is little chance to get the
> precise naming restrictions incorrect: responsibility for selecting
> names is now with the repository operator/software developer, which
> is where I think it belongs.
Hi all,
I've been following this discussion with interest since the company I'm
contracting with does a fair bit of cross-platform development.
Obviously we've hit the case-sensitivity issue and we've dealt with that
by renaming the file so that it doesn't run into issues under NTFS.
This has worked well with the understanding that the build system used
in building the software needs to be changed as well to reflect the name
change. Not always an easy feat.
I'm sure it's possible to have subversion do "something" with the file
names. But what about the build system? Should then make, rake, scons,
ant, nant etc understand what to do if it can't find a file or directory
that it's expecting?
So far for us, by far the easiest thing to do has been to rename the
file to something that doesn't conflict on the target platform.
Just my two cents,
Rick
- application/x-pkcs7-signature attachment: smime.p7s
Received on Mon Feb 6 23:36:42 2006