[systemd-devel] [PATCH v2] libudev: fix symbol version for udev_queue_flush() and udev_queue_get_fd()

[Zbigniew Jędrzejewski-Szmek]

On Tue, Feb 03, 2015 at 07:29:06PM +0100, Michael Biebl wrote:
> 2015-02-03 15:18 GMT+01:00 Kay Sievers <kay at vrfy.org>:
> > On Tue, Feb 3, 2015 at 3:09 PM, Zbigniew Jędrzejewski-Szmek
> > <zbyszek at in.waw.pl> wrote:
> >> On Tue, Feb 03, 2015 at 01:34:23PM +0100, Lennart Poettering wrote:
> >>> On Thu, 29.01.15 15:00, Zbigniew Jędrzejewski-Szmek (zbyszek at in.waw.pl) wrote:
> >>>
> >>> > On Sat, Aug 30, 2014 at 11:56:30AM +0200, Kay Sievers wrote:
> >>> > > On Sat, Aug 30, 2014 at 2:09 AM, Michael Biebl <mbiebl at gmail.com> wrote:
> >>> > > > Updated patch which the correct version information.
> >>> > >
> >>> > > Applied.
> >>> >
> >>> > Hm, I think this was an unintentional ABI break. udev_queue_flush @@ LIBUDEV_183
> >>> > was removed, udev_queue_flush @@ LIBUDEV_215 was added. The linker cannot know
> >>> > that this is the same symbol.
> >>>
> >>> Hmm can you elaborate on this? I missing the context? Is there
> >>> something to fix here?
> >> For two releases systemd had a symbol, which then got removed.
> >> Anyone compiling during that time and using it, would get a crashing
> >> binary after installing systemd from the latest version.
> >>
> >> Although this happened a while ago and nobody complained suggests
> >> that there were few users, and any there were got recompiled anyway.
> >
> > There should be no external users of these symbols, they were just
> > exported for consistency.
> 
> 
> http://codesearch.debian.net/results/udev_queue_flush
> confirms that there are no external users.
> 
> If there are only for internal use, maybe we should consider hiding them.
Please don't. Once they're public, they're public. We can consider doing that
when bumping so version for another reason.

Zbyszek