Web lists-archives.com

Re: [PATCH v7 2/2] json-writer: t0019: add Python unit test

On Wed, Jun 06, 2018 at 09:49:09PM -0400, Todd Zullinger wrote:

> Ramsay Jones wrote:
> [...]
> > I don't run the p4 or svn tests, so ... :-D
> Heh, lucky you. :)
> I try to run them all as part of the fedora builds since
> they cover much more than I'd ever use.  That's the main
> reason I noticed the bare python.  That would trip me up
> when it came time to build on a near-future fedora where
> python isn't installed by default and I only wanted to
> require python3 for the build/runtime scripts.

I'm not actually sure those svn bits involving python are usable by real
users. The main tool that people us for svn interop is git-svn, which is
written in perl.

There was an effort to have a real "git-remote-svn" helper (so that you
could just seamlessly "git clone svn://..." instead of using the "git
svn" wrapper.

So we have the work in vcs-svn, which gets built as part of a regular
compile. But AFAICT it is only used for git-remote-testsvn,
t/helper/test-svn-fe, and contrib/svn-fe. None of which have seen any
substantive work since 2012[1].

I suppose it's possible somebody could be using "git clone testsvn://"
in the wild, but the name would hopefully warn them off. I have no idea
how usable that work is in practice.


[1] Browsing "git log", most of the commits are just tree-wide
    cleanups, or fixing some compilation or dependency error. The last
    "real" commit seems to be around 8e43a1d010 (remote-svn: add
    incremental import, 2012-09-19).

    I wonder if it is worth dropping this experiment as incomplete and
    unmaintained. People have obviously spent time dealing with the code
    for various fixups, but I think the whole thing may essentially just
    be dead code. Or maybe people really are using contrib/svn-fe. Like
    I said, I have no clue if this stuff is even usable, but we
    certainly haven't packaged it to be seen by most users.