Andre Noll [Fri, 13 Sep 2013 05:47:41 +0000 (05:47 +0000)]
build: ldflags conversion: libfaad.
This adds the linker flags for libfaad to LDFLAGS in Makefile.in
for those targets that need it. This avoids some duplication in
configure.ac. No cppflags conversion in this patch.
Andre Noll [Fri, 13 Sep 2013 05:36:22 +0000 (05:36 +0000)]
build: cppflags/ldflags conversion: libspeex.
This fixes the --with-speex-headers option, which has never worked,
by adding the speex_cppflags output variable and including the given
flags in CPPFLAGS for the three objects that need it.
The straight-forward conversion for speex linker flags is also
performed in this patch.
Andre Noll [Fri, 13 Sep 2013 05:24:07 +0000 (05:24 +0000)]
build: ldflags conversion: libogg.
The next step on the way to the removing the per-executable ldflags
variables. This sets ogg_ldflags in configure.ac, creates an output
variable from it, and adds it to LDFLAGS in Makefile.in for all
executables that need it.
Andre Noll [Fri, 13 Sep 2013 05:12:58 +0000 (05:12 +0000)]
build: cppflags/ldflags conversion: libid3tag
This removes the rule for mp3_afh.o so that this object will instead
be created by means of the generic rule. To make this work, CPPFLAGS is
modified for this target to include suitable cpp options for libid3tag
as determined during configure.
Similarly, the linker flags for libid3tag are removed from
server_ldflags, afh_ldflags, play_ldflags, recv_ldflags and
play_ldflags, and appropriate linker flags are added to LDFLAGS for
all executables that need it.
Andre Noll [Fri, 13 Sep 2013 03:13:38 +0000 (03:13 +0000)]
build: Remove duplicates from $all_objs.
In Makefile.in $all_objs is defined as the concatenation of the
object lists of all executables. This variable is used to determine
the set of dependency files to include. Since the concatenation
contains common objects more than once, we currently include common
dependencies multiple times.
This is not a problem, but it may slow down the build. The make sort
function removes duplicate words, so by sorting $all_objs we make
sure to include each dependency file only once.
Andre Noll [Fri, 13 Sep 2013 01:41:15 +0000 (01:41 +0000)]
build: Replace error2.pl by error2.c.
The C code is much faster than the old perl script. Since the resulting
executable runs on the build system, the compiler for this system
must be called to compile error2.c. The new HOSTCC variable takes
care of this.
With this patch applied, the only remaining dependencies on perl are
help2man and autoconf.
Andre Noll [Sun, 8 Sep 2013 01:11:16 +0000 (01:11 +0000)]
build: Separate curses detection from gui configuration.
It's cleaner to separate feature/library detection from the logic
that decides whether an executable can be built.
This patch creates a separate section for para_gui which contains
the list of all objects for para_gui and sets $build_gui to either
"yes" or "no", depending on whether a usable curses library was found.
Andre Noll [Fri, 6 Sep 2013 01:39:20 +0000 (01:39 +0000)]
build: Move down fade section.
The logic which checks whether an executable can be build on the
given system should come after the tests for features and libraries.
The section for the para_fade executable violates this rule, so move
it down to where it belongs.
Andre Noll [Sun, 8 Sep 2013 01:10:56 +0000 (01:10 +0000)]
build: Add object "cmdline" suffix only once.
This is equivalent, but it is both simpler and less error-prone than
calling add_cmdline() for each check. Change all affected assignments
to the one-line-per-object format while we're at it.
Andre Noll [Fri, 30 Aug 2013 05:02:06 +0000 (05:02 +0000)]
build: Fix make install for cross builds.
make install discards symbols from executables by using the -s option
to the install command. If no install command was detected at configure
time, the install-sh script shipped in the paraslash source tree is
used instead.
This is currently broken for cross-builds because install -s runs the
native strip command by default but the cross-strip is required here.
Both the install script and the install command from gnu coreutils
support custom strip commands, but the syntax for specifying these are
different. A decent install program is available at least on FreeBSD,
NetBSD, MacOS and Linux, so let's just kill the script and use the
--strip-command option of the install program to allow the user to
specify a suitable cross-strip.
Andre Noll [Fri, 30 Aug 2013 00:10:41 +0000 (00:10 +0000)]
build: Generate man pages directly from ggo files.
Currently all man pages are created by help2man which executes the
para_xxx binaries just compiled with the --detailed-help and --version
options given to obtain the input for the man page to create.
The obvious shortcoming of this approach is that it simply does not
work when cross-compiling. There is another disadvantage though: Since
the man page file (para_xxx.1) depends on the executable (para_xxx),
any code change causes all affected man pages to be recreated, even
if nothing has changed that would alter the man page content.
The good news is that gengetopt can create the help output without
generating or compiling any code. The bad news is that help2man
insists on executing a program with --help and --version to get the
output.
This commit teaches Makefile to create a dummy shell script for each
executable which accepts the above options and runs gengetopt on the
.ggo file to obtain the help text.
This makes cross-compiling possible and shortens rebuild times since
with the patch applied, the man page is only recreated when the ggo
file changes.
Andre Noll [Sat, 31 Aug 2013 04:19:53 +0000 (04:19 +0000)]
build: Remove fnmatch, malloc, realloc tests.
These cause more harm than good. When cross-compiling for the arm-based
rpi, weird rpc_malloc link failures showed up. Removing the tests
solves the problem.
Andre Noll [Thu, 29 Aug 2013 22:57:11 +0000 (22:57 +0000)]
build: Make crypto optional.
Now that all three executables that need a crypto library have been
made optional, it is no longer a fatal error if no library was found.
This patch makes the build succeed in this case, albeit para_server,
para_audiod and para_client will not be built.
Andre Noll [Thu, 29 Aug 2013 21:42:43 +0000 (21:42 +0000)]
build: Make audiod optional.
The previous two commits made para_server and para_client depend
on crypto support. This patch does the same thing for para_audiod.
It adds a section that sets $build_audiod to yes or no, depending
on whether a crypto library was detected. This clears the path to
making crypto support optional.
Andre Noll [Thu, 29 Aug 2013 21:09:45 +0000 (21:09 +0000)]
build: Build para_client only if crypto lib was found.
This creates a separate section for para_client, similar to the one
for para_server that was added in the previous patch.
The new section sets $build_client to either yes or no, depending on
whether a crypto library (openssl nor gcrypt) was found by the tests
performed earlier. This test is always true at the moment since we
abort anyway if neither library was detected. But this is about to
change, so..
Andre Noll [Thu, 29 Aug 2013 20:56:33 +0000 (20:56 +0000)]
build: Create a separate section for para_server.
The new section is added after osl and crypto detection has been
performed. We only build para_server if both libosl and a crypto
library was found.
This check is not necessary at the moment since we abort earlier if
neither openssl nor gcrypt was detected, but it will eventually allow
to make crypto support optional.
Andre Noll [Thu, 29 Aug 2013 20:35:54 +0000 (20:35 +0000)]
build: Replace $extras by per-executable variable.
This changes the summary at the end of the configure script to
output one line per optional executable rather than to list all
optional executables in a single line.
The new build_xxx variables are also useful to compute the set of
all $all_errlist_objects at a single location, which will be done
in a subsequent patch.
Andre Noll [Sat, 14 Sep 2013 05:34:54 +0000 (05:34 +0000)]
build: Remove two dead recipes from Makefile.in.
In commit f5fb7ab5 we switched from skencil to dia for creating the
overview.pdf file. This commit missed to remove the Make recipes
for creating the pdf file from the skencil sources. Remove it now.
Andre Noll [Thu, 29 Aug 2013 19:09:18 +0000 (19:09 +0000)]
build: Remove check for buggy snprintf().
This check has not found any problematic snprintf() so far. But more
importantly, it uses AC_RUN_IFELSE() which should be avoided since
it can not work for cross-compile setups. Remove the check.
Andre Noll [Wed, 4 Sep 2013 20:14:12 +0000 (20:14 +0000)]
mood: Deduplicate score formula.
mood.c computes the score of an audio file from its subscores (mood,
last_played, num_played) at two locations, both of which use the same
formula (a + b + c) / 3.
Get rid of this duplication by doing the calculation only once in
compute_dynamic_score(). Rename this function to compute_score() as
it now includes the (static) mood score as well.
Andre Noll [Wed, 4 Sep 2013 19:26:46 +0000 (19:26 +0000)]
mood: Combine compute_num_played_score() and compute_last_played_score().
These two functions have only one caller, compute_dynamic_score(), which
just adds the two return values. Let's combine all three functions to a
single one.
Andre Noll [Wed, 4 Sep 2013 19:18:59 +0000 (19:18 +0000)]
mood: Don't add files without valid information to the score table.
Currently if get_afsi_of_row() fails, we add the file to the score
list anyway. This function should never fail, and if it does, we are
in serious trouble anyway and should return the proper error code
rather than success.
Andre Noll [Sun, 6 Oct 2013 16:05:27 +0000 (18:05 +0200)]
Merge branch 't/audiod_cleanups'
Was cooking for ~2 weeks
1ff4b2 audiod: Don't compute stat info unnecessarily.
d80f5d audiod: Force status dump every 5 seconds.
e34e5f audiod: Improve status timeout handling.
fe7442 audiod: Reorder exit log messages a bit.
32a078 audiod: Fix memory leak on exit: stat client.
706778 audiod: Fix memory leak on exit: gengetopt.
eec23a audiod: Fix memory leak on exit: close slots.
cd5f0a audiod: Unify startup messages of receivers, filters and writers.
597966 audiod: Print a warning if filters are ignored.
a94e8e audiod: Remove pointless continue statement in com_stat().
f45bc6 audiod: Make command array static.
c96792 audiod: Fix error message on status errors.
Andre Noll [Sat, 28 Sep 2013 14:27:56 +0000 (16:27 +0200)]
build: Fix --with-opus-libs.
Due to a silly cut & paste bug this option sets $speex_libs rather
than $opus_libs. Since the assignment comes after speex detection,
the option is effectively a no-op.
Andre Noll [Sat, 30 Mar 2013 21:30:54 +0000 (21:30 +0000)]
btr: Simplify btr_node_status().
This changes btr_node_status() to check for errors before looking
at queue sizes. In certain cases this avoids to call the possibly
expensive btr_get_output_queue_size().
If no more input is going to arrive for an internal node whose output
queue is full, btr_node_status() now returns EOF immediately, which
is better than the previous scheme where we waited for the output
queue to become empty before returning EOF.
Andre Noll [Sun, 31 Mar 2013 15:34:50 +0000 (15:34 +0000)]
buffer_tree: Improve btr_splice_out_node().
This changes btr_splice_out_node() to take a pointer to a btrn, just
like btr_remove_node(). This allows to set the variable to NULL after
the node has been spliced out. The callers are updated accordingly.
Andre Noll [Thu, 22 Aug 2013 21:13:37 +0000 (23:13 +0200)]
crypt: Add workaround for non-fork-safe PRGs.
Some PRNGs implementations suffer from the problem that after a fork()
the PRNG state of parent and child process differ only by the child pid
which is mixed into the state. Certain versions of openssl are known to
contain this flaw.
On such implementations two command handlers will generate the same
challenge and session keys if their pid is identical. This may happen
due to pid wrapping.
This patch works around this shortcoming by reading some pseudo random
bytes in the parent process after each fork().
Andre Noll [Sun, 22 Sep 2013 14:45:10 +0000 (16:45 +0200)]
play: Fix prev command.
The command handler for com_prev() stores the number of the
previous (valid) file in ->next_file of struct play_task and sets
->rq to CRT_FILE_CHANGE. The real work is supposed to be done in
load_next_file() which is called from the main post_select() function.
However, load_next_file() ignores ->next_file and computes the number
of the next file itself. Hence the "prev" command acts exactly as
"next", i.e. it selects the next rather than the previous file.
Fix this by ignoring ->next_file only if ->rq is CRT_NONE, which
indicates a normal end of file condition.
Andre Noll [Sun, 7 Jul 2013 12:06:01 +0000 (14:06 +0200)]
audiod: Reorder exit log messages a bit.
On exit only one emergency message should be printed. If para_audiod
receives a signal which causes it to exit, we currently print two
messages. This downgrades the loglevel of the first message to NOTICE,
and moves the emergency message to the bottom of clean_exit().
Andre Noll [Sun, 7 Jul 2013 12:02:40 +0000 (14:02 +0200)]
audiod: Fix memory leak on exit: stat client.
Currently we don't close stat clients on exit which results in (benign)
memory leaks. Introduce the new public close_stat_clients() and call
it from clean_exit() to avoid the leak.
This patch also removes the pointles local variable "fd".
Andre Noll [Sat, 6 Jul 2013 20:26:03 +0000 (22:26 +0200)]
audiod: Fix memory leak on exit: close slots.
Currently we don't bother to close slots on exit. This is no problem
but it causes valgrind to report a bunch of memory leaks. This patch
makes it close all writers, filters and receivers on exit.
To this aim, the cleanup part of close_unused_slots() is abstracted out
into the new close_slot(), which is now also called from clean_exit()
for each slot, just before para_audiod exits. In order to avoid
forward declarations, clean_exit() had to be moved below the two
other functions.
Andre Noll [Thu, 5 Sep 2013 04:18:48 +0000 (04:18 +0000)]
audiod: Unify startup messages of receivers, filters and writers.
During buffer tree startup para_audiod logs one message with log
level NOTICE for the receiver and one for each filter. The writer
startup message is printed only with loglevel INFO though.
Upgrade the loglevel of this message to NOTICE as well.
Andre Noll [Mon, 2 Sep 2013 01:46:32 +0000 (01:46 +0000)]
audiod: Print a warning if filters are ignored.
If the regular expression of a filter specifier does not match any
audio formats, we ignore this filter spec silently. Most likely this
was not intended and deserves a log message, so let's add it.
Andre Noll [Sat, 20 Jul 2013 10:28:18 +0000 (12:28 +0200)]
audiod: Fix error message on status errors.
If para_audiod can not parse a status item received from para_server,
it currently prints
task_notify: notifying task client: audiod switched off
btr_remove_node: removing btr node client from buffer tree
unregister_task: unregistering client (audiod switched off)
This is misleading at best. The error message for status timeouts has
the same problem. This patch makes para_audiod print meaningful error
messages in both cases.
Andre Noll [Wed, 4 Sep 2013 04:35:38 +0000 (04:35 +0000)]
configure: Print opus audio file handler if opus lib was found.
The variable is called $audio_format_handlers rather than
$server_audio_formats. The value is only used to display the supported
audio formats, so this bug is rather benign.
Andre Noll [Fri, 6 Sep 2013 23:21:44 +0000 (23:21 +0000)]
btr: splice-out fix.
When splicing out a node we set the ->parent pointer of all child
nodes to the parent of the given node, and move each child to the
->children list of the parent.
Except when there is no parent. If the given node was a root node
(or an internal node whose parent vanished), we leave the ->children
list untouched. In this case the assertion a few lines later triggers
and aborts the program. Fix this by removing the nodes from the list.
Such nodes have become root nodes themselves, so they should not be
on any list of children.
Andre Noll [Sun, 1 Sep 2013 17:48:46 +0000 (17:48 +0000)]
fecdec: Avoid fecdec output buffer overruns.
The size of the buffer tree pool of the amp filter is 64K, which
is smaller than BTRN_MAX_PENDING of buffer_tree.c (96K). The latter
value is used in btr_node_status() as follows. If input for a buffer
tree node is available and there is less than BTRN_MAX_PENDING bytes
in the output queue of the node, the function returns 1 to indicate
that the node should continue to process its input.
This can result in a fatal error condition when the buffer tree pool
fills up completely. Avoid this by increasing the pool size to 128K.
Andre Noll [Sun, 1 Sep 2013 15:54:25 +0000 (15:54 +0000)]
vss: Avoid assertion "i == g->num_header_slices - 1".
This fixes a very old bug in the timing code for fec slices. Depending
on the number of bytes to submit and on the number of slices of the
current fec group it may happen that we reserve more slices than
needed for the audio file header. This is not a problem per se, but the
above assertion hits in this case and aborts para_server.
Fix this by removing the assertion, and make sure that all reserved
header slices are properly initialized.
Andre Noll [Thu, 29 Aug 2013 22:20:30 +0000 (22:20 +0000)]
configure: Fix restoring LDFLAGS.
The configure script saves the values of CPPFLAGS, CFLAGS and LDLFAGS
before each check and restores the old values afterwards. However,
for LDLFAGS this was broken at several places, since we used LD_FLAGS
(with an underscore) for saving the value and LDFLAGS for restoring it
(or vice versa).
This bug had the effect that a user-specified value of LDFLAGS was
ignored in some tests because LDFLAGS was restored to the empty
value LD_FLAGS during an earlier test.
This patch changes all occurrences of LD_FLAGS in configure.ac to
LDFLAGS.
Andre Noll [Thu, 29 Aug 2013 23:34:13 +0000 (23:34 +0000)]
Don't include $codename in ggo version string.
This variable no longer used. Since we do not rely on gengetopt's
implementation of --version any more, this does not matter much.
Let's fix it anyway.
Andre Noll [Tue, 23 Jul 2013 07:23:32 +0000 (09:23 +0200)]
Avoid gcc warning on FreeBSD.
FreeBSD 9.1 ships a patched gcc-4.2.1 which complains about use of
an initialized variable:
fd.c: In function 'for_each_file_in_dir':
fd.c:728: warning: 'dir' may be used uninitialized in this function
This warning is bogus since para_opendir() only leaves the dir
pointer unset on failures. But in this case for_each_file_in_dir()
returns early without using the pointer.
This commit changes para_opendir() to set dir to NULL before doing
anything else. This avoids the warning at almost no cost.
Andre Noll [Sat, 6 Jul 2013 14:01:26 +0000 (16:01 +0200)]
extract_v4_addr(): Don't return a structure.
When compiling with -Waggregate-return, gcc issues the following warning:
net.c: In function 'extract_v4_addr':
net.c:682: warning: function returns an aggregate
acl.c: In function 'acl_check_access':
acl.c:60: warning: function call has aggregate value
It's not entirely clear how bad it is these days to return a structure, but
extract_v4_addr() is the only function in the tree which does it. Since it
can be easily changed to take a pointer instead and there is only a single
caller, let's implement it this way.
Andre Noll [Sat, 6 Jul 2013 14:23:31 +0000 (16:23 +0200)]
Make log functions static.
All PARA_XXX_LOG() macros use the public para_log function pointer
for quite some time now, so the actual implementations of the various
log functions can be made static.
Andre Noll [Wed, 4 Sep 2013 21:53:32 +0000 (21:53 +0000)]
GIT-VERSION-GEN: Don't include output of git update-index.
GIT-VERSION-GEN calls git update-index which might write something
like "configure.ac: Needs update" to stdout. We don't want this text
to be part of the version string, so make sure we discard the output
of git update-index.
Andre Noll [Fri, 23 Aug 2013 17:35:56 +0000 (19:35 +0200)]
Merge branch 't/web_improvements'
The new pages have been tested on slnx with no problems, so they're
ready for prime time.
329719 web: Move menu down a bit.
c6ac6c web: Make image rule global.
c53634 Replace main page by NEWS.
3c2dfc NEWS: Replace "to be announced" text.
c2ea0c web: Add tarball and signature link of recent release to NEWS.
b35afc web: Adjust sizes of monospace fonts.
7fb28a web: Avoid gray area on gitweb page.
f3f978 web: Move important parts of the style sheet to the top of the file.
57058e web: Remove some duplicate rules in para.css.
5005fb web: Combine the two .css files.
d74c89 web: Add header to gitweb and API reference sites.
076b70 web: Increase size of log messages and diffs.
0384fd web: Unify link colors.
893669 web: Fix gitweb link on doxygen pages.
607d79 web: Unify <br> elements.
40ceca web: Remove valign tag in header files.
37bfc1 web: Further unify header.html, header2.html and gitweb_header.html.
197add web: Unify css rules for tables.
21422a web: Remove horizontal line at the top of the page.
53963d web: Remove style element in html.
4c061d web: Make the body selector of the two css files consistent.
f9fadc web: Use lower-case css selectors.
22fb31 web: Increase font size.
2f201d web: css whitespace fixes.
51445c web: Remove commented-out sections of css files.
71d445 web: Remove border="0" tag.
02c1a5 web: Don't set border, cellpadding and cellspacing.
6e0b01 web: Remove basefont tag.
42a4cd Update README.
2210c6 web: Remove explanation of the paraslash acronym.
Andre Noll [Fri, 23 Aug 2013 08:57:38 +0000 (10:57 +0200)]
overview.pdf: Fix overfull lines.
The Ubuntu version of dia (which is used to create the public file
on the web page) seems to behave differently than the self-compiled
version used on my development box.
This resulted in a rather strange looking pdf. Fix this by manually
adjusting lines using the Ubuntu version.
Andre Noll [Thu, 20 Jun 2013 21:20:13 +0000 (23:20 +0200)]
Change all multi-word options to dashes instead of underscores.
Currently some of these options use dashes to separate the words
while others use the underscore character instead. The auto-generated
--detailed-help option contains a dash, so the only consistent way
is to use dashes everywhere.
Hence this commit replaces all underscores in option names by dashes
and updates the documentation accordingly.
Fortunately, this does not require to change the code, because
gengetopt translates dashes into underscores in the generated C code.
It is, however, an API change, so this is 0.5.0 material.