Strange commit

[Ulrich Spörlein]

Am Do., 2. Apr. 2020 um 21:35 Uhr schrieb Ed Maste <emaste at freebsd.org>:
> On Thu, 2 Apr 2020 at 14:51, Kyle Evans <kevans at freebsd.org> wrote:
> > I'm not sure I understand what's going on here... it's a merge commit
> > that didn't have any new files or moved files. This is the exact
> > result you get from svn merge -c from the vendor/ branch with just
> > file modifications -- note the svn:mergeinfo.
>
> My working theory right now is that svn2git fails to detect merge
> commits that only modify existing files and do not copy any new files.
> I found a document[1] that svn2git should parse mergeinfo, but that's
> in the context of a working / topic branch, not a vendor branch. That
> document does point out that "grafts" can be used to retrofit the
> history after the conversion, if we must.
>
> More investigation is needed.
>
> [1] https://web.archive.org/web/20190515003725/http://blog.agavi.org/post/16865375185/fixing-svn-merge-history-in-git-repositories

Can folks help out a bit on how merginfo is stored and whether it is
actually stored on the revisions?

svn prophet svn:mergeinfo -r337810 is not very useful output (and
doesn't say anything about /vendor/lua !)

I haven't fully read through
http://subversion.apache.org/docs//api/1.11/svn__mergeinfo_8h_source.html
but it talks about mergeinfo being on paths (which is also my faint
recollection)

Cheers
Uli