Re: svn commit: r234684 - head/lib/libusb
- To: Robert Millan <email@example.com>
- Cc: firstname.lastname@example.org
- Subject: Re: svn commit: r234684 - head/lib/libusb
- From: Hans Petter Selasky <email@example.com>
- Date: Thu, 26 Apr 2012 23:55:46 +0200
- Message-id: <firstname.lastname@example.org>
- In-reply-to: <CAOfDtXO48tkN_-kUUZwU+RxkUyPHBVzrxmBUO8rGnUi3+Y1Lsg@mail.gmail.com>
- References: <email@example.com> <CAOfDtXO48tkN_-kUUZwU+RxkUyPHBVzrxmBUO8rGnUi3+Y1Lsg@mail.gmail.com>
On Thursday 26 April 2012 19:54:00 Robert Millan wrote:
> El 25 d’abril de 2012 23:20, Hans Petter Selasky
> <firstname.lastname@example.org> ha escrit:
> > Isn't libusb2debian a binary compatible drop-in replacement for
> > applications which are compiled with the linux libusb.
> By linux libusb do you mean the one from http://www.libusb.org/ ?
> AFAIK this one is a portable library, at least it builds on
> GNU/kFreeBSD too.
The one from libusb.org does not work under FreeBSD.
> I wasn't directly involved in this episode. We're providing both
> (libusb from libusb.org and libusb from FreeBSD) but I'm not sure why.
> I suppose the one from FreeBSD has better support for kernel of
> FreeBSD? Or perhaps because they're not API-compatible and
> applications could need either of them?
> > If applications are compiled from
> > source there is no problem, though I see some applications like
> > python-libusb that hardcode structure sizes.
> That sounds like a problem. Could you be more specific? (and put
> email@example.com on CC)
> > Just leave the libusb2debian as-is then, and we'll have this in once the
> > 10-version is out.
> Ok, but if you changed ABI in 10-current, then please do bump the
> soname. We have a huge amount of trouble every time an ABI bump is
> missed (this happened with libgeom a while ago).
Yes, this is done now.