1. Don't define _XOPEN_SOURCE, at least with GCC. As ALWAYS, this macro undefines
a lot of the symbols we need. As of libshout 2.1, this define will be moved
into the "if not $GCC" category so we don't have to deal with it so often.
2. Irix defines inet_pton but not sockaddr_storage. We could fake it by using
an in6 structure, but why bother? Just don't check for inet_pton unless
we have struct sockaddr_storage.
svn path=/trunk/m4/; revision=5087
I suspect all this may be unnecessary, only because I've never heard of this
before. But then, I don't do much library work, and better safe than sorry.
svn path=/trunk/net/; revision=5083
I removed icy- support from the generic send headers function, as it is my
understanding that only MP3 sources use icy headers. PLEASE correct me
if I am wrong.
I also added code in the mp3 function to convert ice-audio-info bitrate to
an icy-br header. ice-audio-info is for YP, icy-br for clients. Perhaps we
should send both to clients though?
svn path=/trunk/icecast/; revision=5079
Basically we define _mangle() in config.h, and use that in timing.h (if available) to prepend the package name to all the symbols in the header and including C files. It's thus fairly invisible.
Unfortunately the prefix is hardwired, so it's another thing that has to be maintained by hand (not that is changes often, but we can't make it a macro). AH_VERBATIM seems to be the only way to add a macro that takes arguments, and it explicitly doesn't do shell expansion. Further, @PACKAGE@ substitution doesn't happen on config.h.in. Insert standard autotools frustration here.
svn path=/trunk/timing/; revision=5060
causing stuttering in some clients. Thank goodness icecast uses an
odd metainterval, or this bug would never have been discoverd :)
svn path=/trunk/icecast/; revision=5044
- all yp adds/touches are now done within a single thread.
This should eliminate any "thread growth" issues and make
things behave much nicer when yp problems arise. We should
eventually change the add/touches to non-blocking sockets,
which will be needed for large numbers of streams doing adds/touches.
svn path=/trunk/icecast/; revision=5012
- most of the yp processing now moved into the yp add/touch thread
- use ice/icy-public rather than ice/icy-private to coorespond to libshout and general convention
- memory leak fixed with audio-info
svn path=/trunk/icecast/; revision=5009
even if action-if-not-found is defined.
Rename have_shout to xt_have_shout, in conformance with macro writing guidelines :)
svn path=/trunk/m4/; revision=5006
1. src/net tests abstracted into XIPH_NET macro. We should do the same for
src/thread, but it's a little more complicated - it will have to
interact with the caller (eg some apps may want --disable-threads)
2. uintxx_t, socklen_t tests encapsulated, moved to xiph_types.m4
3. solaris fixes for socket func, nanosleep detection
4. added $LIBS to SHOUT_LIBS for *-config scripts. Good or bad idea?
We'll definitely need a libshout beta 2.
svn path=/trunk/m4/; revision=5005
ATTN Mike!
make install now installs the xsl files in pkgdatadir ($prefix/share/icecast) web
and admin directories, and builds icecast.xml accordingly. icecast.xml is now
installed in $sysconfdir/etc, and IMHO icecast should attempt to find a config
file there, and only demand one on the command line if it can't.
svn path=/trunk/icecast/; revision=4956
It relies on pkg-config or shout-config (in that order) for compiler/linker
flags. --with-shout-prefix is gone. Also gone is --disable-shouttest. I don't
understand the advantages of that home built script over the standard header/
func checks, so I used those. I guess AC_TRY_RUN isn't invoked, but I feel
(mildly) that the business of autoconf is to ensure the program builds
correctly, not runs correctly. Run errors can usually be fixed after the build
(by tweaking LD_LIBRARY_PATH, ld.so.conf, crle etc) without rebuilding.
I've ported ices 0 to this macro in order to test it, but haven't put the various
fallbacks through the ringer yet. I will before release.
TODO: I don't think this file is being included with make dist. Should it be?
Probably. Should it be installed? Probably, but less so.
svn path=/trunk/m4/; revision=4939
