On Tue, Nov 19, 2019 at 12:04 PM Daniel Shahaf <d.s_at_daniel.shahaf.name>
wrote:
> Nathan Hartman wrote on Tue, 19 Nov 2019 04:53 +00:00:
> > Moving this from users@ to dev@...
>
> [Aside: when doing this, it's a good to post a short response on users@
> saying the conversation continues on dev@, for anyone who wants to
> follow along but isn't subscribed to both lists.]
>
> > On a somewhat related note (and at risk of blasphemy), if we want to
> > provide a way for users to know what revision their svn install was
> > built from, which would work for those who build from tarballs also,
> > would it make sense to create a header file for this purpose, turn on
> > svn:keywords for it, and put "$Revision: $" into a string
>
> If we do this, we'll run into this issue:
>
> http://subversion.apache.org/faq#version-value-in-source
>
> tl;dr: Putting $Revision$ in a header file will expand to the last
> changed revision of _that_ header file, not to the last changed revision
> of anything in cwd.
>
Ah, yes. Thanks for pointing that out. That's not what we want.
The concept of having the binary know the `svnversion` info of the
> working copy it was built from does make sense, however. We could make
> gen-make.py add «-DFOO=`svnversion`» to CFLAGS. (There's already
> something like that, «fakedefines», in the Windows-specific part of the
> build generator.)
>
That wouldn't do what we want either. If you build from a tarball,
it'll bake "Unversioned directory" into the binaries. Also it would
mean that you must have svn installed in order to build svn,
regardless whether the sources were obtained from distribution tarball
or otherwise.
So... I guess what we're doing now makes sense.
Oh well, it was a nice thought.
Nathan
Received on 2019-11-19 19:25:58 CET