2008/11/28 Steve Huston <[EMAIL PROTECTED]>:

>> Obviously that was simple to fix, and I've also added a custom build
>> step (and the spec file) to the project so that a user doesn't have
> to
>> run nmake separately from the command line.
>
> I originally did this, but removed it during the git-to-svn merge. My
> thoughts were that since the Qpid release process generates the files,
> "normal" users won't need to generate them and may not have the
> ruby/python needed to do it. It's only needed by qpid developers
> working from svn, so I moved the generate to a separate procedure. I
> have a Windows-specific install document written as well.

I just looked at the tar.gz produced by Rafi for the M4 beta
(http://people.apache.org/~rhs/qpid-incubating-M4-beta/qpid-incubating-M4-beta.tar.gz)
and it did not contain the generated files?

> Also, even if the generated code is done, there's no way that I've
> found to bring any new files into the project file.

Yes, this is a real pain. It would be nice if VS had an option to add
a directory tree to a project that reflected the current state of the
directory on disk at any point in time.

> I have some things, yes. Additionally, I've been generating the
> sln/vcproj files (and that does find all the generated code) so if you
> check it in, it'll likely show up as much more changed than just what
> you added.

Aha - I agree that generating the project files is the best solution.

> It is not a perfect scheme, and I'm open to suggestion for
> improvements, but adding the generating step is, I think, worse for
> non-developers. Let me know what you think.

I agree that getting users to download both ruby and python just to do
the generation process will be a pain for them (incidentally I don't
know why we need to have two languages involved in code
generation...). If we can get source releases to include the generated
files then I agree that your approach is the right one.

RG

Reply via email to