svn commit: r454171 - head/multimedia/libvpx

[Alexey Dokuchaev]

On Tue, Nov 14, 2017 at 04:12:05PM +0100, Jan Beich wrote:
> Alexey Dokuchaev <danfe at FreeBSD.org> writes:
> > ...
> > => Attempting to fetch https://github.com/android/platform_system_core/commit/b4cff495a1f9.patch
> > fetch: https://github.com/android/platform_system_core/commit/b4cff495a1f9.patch: size unknown
> > fetch: https://github.com/android/platform_system_core/commit/b4cff495a1f9.patch: size of remote file is not known
> > b4cff495a1f9.patch                                    2084  B 6281 kBps 00m00s
> > => Fetched file size mismatch (expected 2066, actual 2084)
> 
> "git format-patch" from Git 2.11 or later honors commit abbreviation.
> No wonder it affects .patch links from GitHub.
> 
> cf. https://github.com/git/git/commit/e6c587c733b4

While it's good that you've found what's causing the issue, it wouldn't
be a problem in the first place if the patch was placed under files/.

> > I don't think that this is more important than being unable to svn up
> > to some past revision and having to play tricks with -DNO_CHECKSUM and
> > hoping everything will be OK and it's just stupid GitHub changed their
> > output again.
> 
> That's optimistic for ports that specify only 7-hexdigits in GH_TAGNAME
> or vanishing MASTER_SITES.

I'm used to deal with vanishing MASTER_SITES; missing distfiles can very
often be found on various mirrors (including stale ones, which is often
a plus), in Debian or Gentoo archives, etc.  This is not the case with
patches, as no one except FreeBSD could be referencing them.

> Not to mention, partial downgrades have never been supported because
> Mk/* bits provide limited backward compatibility.

It's one of very awesome features of our ports, I use it all the time
(as I often prefer earlier, not-so-bloated software versions), and we
should strive to maintain (or at least not deliberately break) it.

./danfe