Signed-off-by: Ralph Giles <giles@mozilla.com>

Implement reverse timestamp calculation for Opus (fixes calculated
timestamps on first page), and correct for pre-skip and end trimming.
Signed-off-by: Ralph Giles <giles@mozilla.com>

Reviewer note: Pretty sure we can make queued_duration overflow
by feeding it a long stream of pages without proper granulepos.
I decided not to worry about it; the caller will get unexpected
nonesense, but the stream is also invalid. Just wanted to note
the issue for future reference.

Cleaner, even if the OS should be nice to us.

That become errors for distcheck ...

unistd.h is no longer included by (as many) other system headers,
and that's needed for read(2).

A void* doesn't fit into int on all architectures.

There have been reports of things breaking for users of very new
autotools now (against libogg and other Xiph libraries), so fix
this one too while it's open in the editor.

Patch from RedDwarf, Closes ticket:1758 (for oggz).

There is more that could be done here, right now the install-data target
installs everything that doxygen generates into $docdir, which with the
current doxygen configuration includes both html and latex docs, the
latter of which aren't actually in a user-presentable form without doing
a further build of them to convert to the desired format.

Probably the html docs should be installed to htmldir, and the latex
docs should not be installed at all in their current form.  If people
do use the latter, then adding top level targets to fully generate
them wouldn't be the silliest thing to do, and then they should be
installed to dvidir/pdfdir/psdir as appropriate.

But so far, being able to alter $docdir is all that was asked for, and
this patch does exactly that.  The rest can wait for someone with the
right itch for now.

Also while we're in here, fix some formatting glitches, some things that
don't follow the normal manpage requirements or font conventions, drop
some formatting markup that isn't actually doing anything, and the note
that it was generated by docbook-to-man which is presumably responsible
for most or all of the above.

Updating the rest of the manpages similarly is left as an exercise for
people/days with more spare time.

Merged from the patch in the Debian packages.  The other half of that
patch was already fixed in e52622ed.

And make a sentence out of "contains multiple Vorbis" while we're here.

Reported by Elrood on #opus.

See mapping of VP8 into ogg spec:
http://people.collabora.co.uk/~slomo/webm/ogg-vp8.pdf

This makes oggz_comment work on opus files.

use read(2) instead of fread(3) in order to avoid stalling when the data comes in slowly (eg, live stream at very low bitrate).

When files with pre-0.9.9 tool names exist in $(exec_prefix)/bin,
"make DESTDIR=${staging_dir} install" tries to delete them, which
may cause a failure for lack of permissions (quite likely to happen
when building a package as a normal user), and is wrong anyway
(nothing outside of DESTDIR should ever be changed by "make install").

With this patch, it will now cd to "$(DESTDIR)$(bindir)" and create
relative symlinks there, if need be. This preverse the old behavior
when DESTDIR=="", except for the fact the symlinks will be relative
rather than absolute.

oggz-sort would miss an input on an iteration in which a stream ends and is
removed from the stream table.

The correction for different starting frame numbers on early Theora streams
was backward, resulting in the timestamp calculation being off by a frame
on all stream version. This was causing oggz-sort and oggz-merge to mis-mux
streams (which oggz-validate was catching properly)

The return value of fopen() simply wasn't being checked.

It was not looking for calc. gpos on packets which did not
 originally have a granule position, nor was it looking for the |
 that's used in split granule positions (a la Theora).

... in oggz-validate; reported by oggkoggk

This fixes an issue with some Kate tracks not playing with liboggplay,
reported by both Viktor Gal and ogg.k.ogg.k

Adapated from patch by ogg.k.ogg.k