Andre Noll [Mon, 24 Apr 2017 18:05:18 +0000 (20:05 +0200)]
Merge branch 'refs/heads/t/xz'
A single patch which modifies the build system to create xz-compressed
tarballs. The patch conflicted against the changes introduced by the
recently merged makefile_conventions topic branch.
Cooking for almost a month.
* refs/heads/t/xz:
Switch to xz to compress tarballs.
Andre Noll [Sun, 23 Apr 2017 19:12:22 +0000 (21:12 +0200)]
Merge branch 'refs/heads/t/dynamic_chunks'
A rather long and intrusive series that finally improves the aac
decoder and audio format handler.
The merge had a minor conflict in error.h, which was easily resolved
by removing both affected error codes.
Cooking for a month.
* refs/heads/t/dynamic_chunks: (24 commits)
afh_recv: Check return value of afh_get_chunk().
aacdec: Decode only one frame per iteration.
aacdec: Combine aac_open() and aacdec_open().
aacdec: Make "initialized" a boolean.
aacdec: Make frame_info local to ->post_select().
aacdec: Remove pointless assignment.
aacdec: Remove pointless check and improve comment.
aacdec: Rename buffer variables.
aacdec: Remove pointless variable p.
aacdec: Remove check which is always true.
aacdec: Kill pointless label.
aacdec: Remove superfluous assignment.
aacdec: Improve and silence error message.
aacdec: Prefer NeAACDecInit() over NeAACDecInit2().
aacdec: Don't eat full buffer on errors.
Combine aacdec and aac_common.
Convert the aac audio format handler to libmp4ff.
aac_afh: Don't create chunk tables any more.
afh: Dynamic chunks.
server: Store max chunk size in database.
...
Andre Noll [Fri, 21 Apr 2017 18:57:38 +0000 (20:57 +0200)]
afh_recv: Check return value of afh_get_chunk().
Starting with commit b6b571e6c (afh: Dynamic chunks) the function
returns an int value and may fail. Although we store the return value
in a variable, the value of the variable is never read, causing the
clang static analyzer to complain (rightfully).
This patch modifies afh_recv_post_select() to fail if afh_get_chunk()
returns negative.
Andre Noll [Wed, 19 Apr 2017 17:13:20 +0000 (19:13 +0200)]
Merge branch 'refs/heads/t/makefile_conventions'
A bunch of changes which strive to make the paraslash build system
more standard.
Cooking for three months.
* refs/heads/t/makefile_conventions:
build: Add target "check" as a synonym for "test".
build: Rename target "tarball" to "dist".
build: Do not strip installed executables by default.
build: Support $(DESTDIR).
build: Rename clean targets.
Andre Noll [Sun, 16 Apr 2017 18:15:35 +0000 (20:15 +0200)]
Merge branch 'refs/heads/t/rm_rc4'
This patch removes support for RC4, making the AES-based stream
cipher mandadory. The aes_ctr128 server feature is made a no-op,
breaking support with very old clients (<= 0.5.1).
Andre Noll [Sun, 9 Apr 2017 11:47:15 +0000 (13:47 +0200)]
Merge branch 'refs/heads/t/rm_asn'
A few cleanups, and one patch which removes an obsolete feature,
getting rid of an open-coded ASN parser.
Cooking for three months.
* refs/heads/t/rm_asn:
Remove unused E_PUBLIC_KEY.
crypt.c: Combine load_key() and get_private_key().
crypto: Remove support for ASN public keys.
crypto: Simplify asymetric key handling.
crypto: Rename check_key_file() -> check_private_key_file().
gcrypt.c: Always initialize result pointer in get_private_key().
Andre Noll [Sun, 3 Jul 2016 14:47:13 +0000 (16:47 +0200)]
Switch to xz to compress tarballs.
xz compresses better than bzip2, and decompression needs less time. It
is also widely used by many other software projects, so there should
be no disadvantage.
We keep removing the .bz2 tarball on make clean, and also keep the
paraslash-*.tar.bz2 entry in .gitignore for the time being.
Andre Noll [Sat, 28 Jan 2017 17:02:33 +0000 (18:02 +0100)]
wmadec: Use read_u32_be().
The shift operation in show_bits() was buggy because p[0] is promoted
to int, and the shift p[0] << 24 results in undefined behavior,
causing the sanitizer of gcc to complain:
bitstream.h:40:21: runtime error: left shift of 230 by 24 places cannot be represented in type 'int'
Andre Noll [Sat, 7 Jan 2017 15:12:05 +0000 (16:12 +0100)]
imdct.c: Replace pointless macro PASS.
The macro defines pass(), a function with only a single caller. We
may as well define the function directly, improving readability and
enabling proper syntax colors.
Andre Noll [Fri, 18 Apr 2014 00:12:40 +0000 (00:12 +0000)]
aacdec: Combine aac_open() and aacdec_open().
Both functions are short, and they are only called once at
startup. Thus, there is no reason to spread out the initialization
over two functions, so let's combine them.
Andre Noll [Sat, 24 Dec 2016 16:41:00 +0000 (17:41 +0100)]
aacdec: Improve and silence error message.
Most errors from NeAACDecDecode() can be handled gracefully. This
patch downgrades the severity level of the corresponding log messages
from ERROR to NOTICE and explains the meaning of the three numbers
of the second log message.
Andre Noll [Sat, 31 Dec 2016 01:39:16 +0000 (02:39 +0100)]
aacdec: Prefer NeAACDecInit() over NeAACDecInit2().
The aac decoder tries to parse the length of the decoder configuration
from the esds atom and passes this information to NeAACDecInit2(). If
we can not figure out the length, we fall back to NeAACDecInit(),
which does not receive the decoder length as a parameter. This is
unnecessary because NeAACDecInit2() is only necessary for decoding
mp4 files, while both para_afh and para_server feed demultiplexed
aac frames to the decoder.
Calling NeAACDecInit() unconditionally allows to get rid of the
code which detects and parses the esds atom.
Andre Noll [Sun, 6 Apr 2014 10:13:10 +0000 (12:13 +0200)]
aacdec: Don't eat full buffer on errors.
On decoding errors, if the call to NeAACDecDecode() did not consume
anything from the input buffer, we currently throw away the full
buffer. This does not necessarily improve matters, so let's just eat
one byte and hope that subsequent calls succeed.
Andre Noll [Thu, 17 Mar 2016 20:59:52 +0000 (21:59 +0100)]
Combine aacdec and aac_common.
With the server side switched to libmp4ff, only the aac decoder needs
the public functions in aac_common. This patch moves this file into
aacdec_filter.c, makes the functions static and removes the aac.h
header file which only contained the prototypes of the previously
public functions.
Makefile.real, configure.ac and error.h need small adjustments due
to the removal of aac_common.c.
Andre Noll [Sat, 24 Dec 2016 15:02:44 +0000 (16:02 +0100)]
Convert the aac audio format handler to libmp4ff.
This changes the aac audio format handler to call the primitives
provided by the mp4ff library. This allows to remove the atom parsing
code from aac_afh.c.
After this change para_server, para_afh, and para_recv no longer
depend on aac_common.c and on the mp4v2 library. The autoconf tests
for libmp4v2 are removed from configure.ac. A subsequent commit will
deal with aac_common.c.
Andre Noll [Sat, 24 Dec 2016 15:02:36 +0000 (16:02 +0100)]
aac_afh: Don't create chunk tables any more.
A previous commit activated dynamic chunks for the aac audio format
handler, so the virtual streaming system no longer consults the chunk
table stored in the audio file table of the osl database.
However, the code to open or add an audio file still assumes that
there is a chunk table and bails out if it can't find it. This patch
changes aft.c to do without a chunk table if the audio format handler
supports dynamic chunks. The afh_supports_dynamic_chunks() helper
needs to be made public because of this.
With chunk tables being optional, the ->get_file_info method of the
audio format handler can safely set ->chunk_table to NULL. It still
needs to compute the maximum chunk size though.
Andre Noll [Wed, 21 Dec 2016 23:57:15 +0000 (00:57 +0100)]
afh: Dynamic chunks.
paraslash chunk tables were designed long ago with the idea that the
full audio file, with the exception of a potential header, is going
to be sent to the client. This allows to store a sequence of offsets
as the chunk table. Each chunk is defined as the contiguous region
of the file given by two consecutive offsets.
For most audio formats, however, not every part of the file corresponds
to encoded audio. We work around this on the client side by letting
the filters detect and skip those parts which can not be fed to
the decoder.
This works generally well, but for the aac decoder we have a rather
ugly hack that skips over any non aac decoded data of its input. This
hack was never very reliable, and the concept of dynamic chunks
finally allows to get rid of it.
Dynamic chunks work as follows. Each audio format handler signifies
support by defining the new ->get_chunk method. In this case
afh_get_chunk() no longer consults the chunk table at all but calls
the new method instead in order to obtain a reference to the chunk.
This comes with a certain overhead at runtime because we need to call
into the functions of the mp4ff library (ships together with faad)
rather looking up the offset in the chunk table.
Only the aac audio format handler supports dynamic chunks per this
commit. To keep the patch size relatively small, this commit does not
touch ->get_file_info() of the aac audio format handler. Therefore,
when a new m4a file is added to the database, the aac audio format
handler still creates the chunk table. A subsequent commit will turn
off this unnecessary operation.
The documentation is updated to mention that mp4ff is now required
for the aac audio format handler. The configure script now checks
for the mp4ff header and the library and deactivates aac support if
it was not found.
Andre Noll [Fri, 23 Dec 2016 16:18:21 +0000 (17:18 +0100)]
server: Store max chunk size in database.
This number is needed up-front for the initialization of the fec data
structures. Currently we recompute it from the chunk table each time
the file is opened for streaming.
We can only get rid of the chunk table concept if we tell the VSS by
other means how to obtain this information.
Fortunately there is an unused 4-byte field in the on-disk afhi
structure, which is always zero at the moment. This patch starts to
use this field to store the maximal chunk size.
For backwards compatibility, when the afhi structure is loaded from
disk at stream time, we check if the field is zero and recompute the
max chunk size as before in this case.
Although the maximal chunk size is generally only needed on the
server side, for consistence we expose it though a new status item
along with chunk_tv and friends.
Andre Noll [Wed, 21 Dec 2016 12:58:28 +0000 (13:58 +0100)]
server: Move ->size from mmd to struct vss_task.
This field stores the size of the memory mapping of the current audio
file, if any. No command handler ever changes the value, and the
value is not read from com_stat(), so there is no reason for ->size
to be part of the mmd shared memory area.
Unlike ->size, the pointer to the mapping is stored in struct vss_task.
Given that vss.c is the only file which reads ->map or ->size, it
makes sense to store both values together in the vss task struct.
Both callers pass a buffer to load_chunk_table() to initialize
afhi->chunk_table. The function does not check the size of the passed
buffer but relies on afhi->chunks_total to tell how many values to
copy from the buffer.
In other words, the buffer size gives an upper bound for the number of
chunks in the chunk table, but this bould is not checked. This patch
makes the code more robust by adding a suitable check. To this aim,
load_chunk_table() is changed to receive a pointer to an osl object
(which contains the buffer size) rather than only the buffer.
One caller, load_afd(), initializes the chunk table from a shared
memory area buffer. It does not know or care about the buffer size
so far. We introduce the the new IPC helper shm_size() to let it tell
the size of the area and pass it to load_chunk_table().
Andre Noll [Thu, 22 Dec 2016 01:38:05 +0000 (02:38 +0100)]
para_server: Never read past the end of the chunk table.
If the server command "ls" is executed with the -c option to print
the chunk table, we assume that afhi->chunks_total many chunks are
stored in the osl disk object that represents the chunk table.
This should be true in general, but since we also know the size of
the osl object, it does not hurt to check this bound as well and
break out of the loop if the next read would access memory beyond
the end of the object.
Andre Noll [Sat, 28 Jan 2017 18:33:26 +0000 (19:33 +0100)]
i9e: Restore file status flags on exit.
The i9e subsystem sets the stdin and stdout fds passed to i9e_open()
to nonblocking mode but misses to restore the original flags in
i9e_close(). This causes terminal applications like dialog to fail
if they are started in the same terminal after e.g. para_play was
executed.
This commit modifies i9e_open() to fetch and save the file status
flags before setting the O_NONBLOCK flag, and i9e_close() to restore
the original value. STDERR is not affected.
Andre Noll [Mon, 18 Jul 2016 21:35:48 +0000 (23:35 +0200)]
Remove compatibility code for abstract unix domain sockets.
Abstract sockets were introduced in paraslash-0.5.5 (2015-09-20). For
backwards compatibility, create_local_socket() was modified to create
an abtract socket *and* an ordinary pathname socket so that old audiod
clients (which connect to the pathname socket) would still work.
This patch breaks compatibility by going back to a single socket,
either abstract (Linux) or pathname (all other systems), as determined
in the first call to create_local_socket(). This allows to drop the
"abstract" argument of init_unix_addr() and the "mode" argument of
create_local_socket().
The code in net.c and the callers get quite a bit shorter since it
has to deal with only a single file descriptor and one type of socket.
Andre Noll [Mon, 22 Aug 2016 15:36:05 +0000 (17:36 +0200)]
build: Add target "check" as a synonym for "test".
make check seems to be more common than make test, and "check" is the
recommended name according to the GNU standards. This patch teaches
the build system to support both targets. It also marks the test-clean,
check and test targets as phony.
Andre Noll [Mon, 18 Jul 2016 16:49:31 +0000 (18:49 +0200)]
build: Do not strip installed executables by default.
The "install" target currently strips all executables while installing
them, which is not recommended according to GNU conventions. Instead,
there should exist the "install-strip" target for this purpose.
Moreover, the current install target assumes the GNU variant of
the install utility because we call install with --strip-program,
an option which is not available on *BSD.
This patch addresses both issues. It removes the autoconf check and
lets the user directly define the path to the install executable by
setting INSTALL, INSTALL_PROGRAM, INSTALL_DATA, as recommended by the
GNU project. These variables are used in the commands to be executed
when the install target is made. They have reasonable and portable
defaults, so not setting them at all should be fine on all supported
platforms, addressing the issue with --strip-program.
The new phony target install-strip runs install -s but does not try
to be smart about how to tell the install implementation which strip
program to use.
The cross compiling example can be simplified to reflect the fact
that CROSS_COMPILE is no longer used in the Makefile.
Andre Noll [Sun, 17 Jul 2016 19:27:12 +0000 (21:27 +0200)]
build: Support $(DESTDIR).
The DESTDIR feature is orthogonal to the existing --prefix option
to configure, and the GNU make manual strongly recommends to support
the feature, so this patch implements it.
The feature works as follows. The content of the DESTDIR make variable
is prepended to each installed target file. It is not set at all in
the Makefile, so files are installed into their expected locations by
default. For example, with the patch applied, one may say
make DESTDIR=/tmp/stage install
to install in /tmp/stage/usr/local.
Other than that, specifying DESTDIR has no effect and the value is
not included in any file contents.
Andre Noll [Tue, 12 Jul 2016 19:56:13 +0000 (21:56 +0200)]
build: Rename clean targets.
The documentation of gnu make recommends to have targets called
mostlyclean, clean, distclean and maintainer-clean while we have clean,
clean2, distclean and maintainer clean.
This patch adds mostlyclean and removes clean2. The former target
removes the executables and all object files, clean additionally
removes the whole build directory and distclean removes, on top of
that, all files created by autoconf. Finally, maintainer-clean removes
the tarball and the files created by doxygen and global.
Andre Noll [Wed, 24 Aug 2016 13:12:53 +0000 (15:12 +0200)]
crypt: Remove RC4 support.
Multiple vulnerabilities have been discovered in the RC4 stream cipher,
rendering it insecure. paraslash stopped using RC4 as the default
stream cipher since version 0.5.2 (2014-04-11), but server and client
still supported the broken cipher for backward compatibility. This
commit removes the compatibility code from both the openssl and
the libgcrypt code base, leaving aes_ctr128 as the only remaining
stream cipher.
The server still announces the aes_ctr128 feature, although it is now
mandatory because the server will enable aes_ctr128 unconditionally,
no matter whether it was requested by the client or not. The client,
on the other hand, still requests this feature, regardless of whether
it was announced by the server or not. This keeps unpatched clients =>
0.5.2 working with new servers and vice versa.
Regarding the public crypto API, sc_new() loses its boolean use_aes
parameter. Otherwise the API remains the same.
The patch also rewrites the crypto section of the manual to not
mention RC4 any more.
Andre Noll [Sun, 10 Jul 2016 14:15:37 +0000 (16:15 +0200)]
gcrypt: Remove open-coded OAEP padding.
The open-coded OAEP padding implementaton of gcrypt.c was necessary to
support old libgcrypt versions which do not have the oaep flag. These
days, everybody ought to have moved on to 1.5.0 (released 2011)
or later, so lets require this version from now on.
decode_rsa() used its key_size parameter only for old gcrypt versions,
so the parameter can be removed. Also, the E_MPI_PRINT error code
has become unused and can be removed from error.h.
Since rsa_decrypt_sexp is now constant, there is no need to have a
variable for this, so we can make it a preprocessor definition instead.
Andre Noll [Tue, 12 Jul 2016 17:54:19 +0000 (19:54 +0200)]
crypto: Remove support for ASN public keys.
These have been deprecated for some years in favor of ssh keys
generated with ssh-keygen(1). Removing support for the deprecated
format allows to get rid of quite some ugly ASN parsing code.
Private ASN keys, however, still need to be parsed in case
libgcrypt is employed as the crypto API. So the parser in
find_privkey_bignum_offset() needs to stay.
Andre Noll [Sun, 10 Jul 2016 20:23:45 +0000 (22:23 +0200)]
crypto: Simplify asymetric key handling.
get_asymmetric_key() and free_asymmetric_key() are public because
para_server maintains a copy to the public key of each user so that
the keys need to be loaded only once. On the other hand, for private
keys (used in para_client) key allocation and freeing is performed
implicitly in priv_decrypt(), and no reference to the key is ever
returned. So the crypto API can be simplified by exposing the interface
only for public keys.
Hence this patch renames get_asymmetric_key() to get_public_key()
and drops the "private" argument. Similarly, free_asymmetric_key()
is renamed to free_public_key().
For public keys the function only called stat(2), which is unnecessary
because only an error from the subsequent open(2) call requires to
fail the operation.
The stat() call is needed for loading private keys though, to make
sure permissions are restrictive enough.
This commit renames the function as indicated in the subject and
drops the second parameter. In crypt.c we now call this function for
private keys only.
Andre Noll [Sun, 10 Jul 2016 20:20:03 +0000 (22:20 +0200)]
gcrypt.c: Always initialize result pointer in get_private_key().
This should not matter since the only caller, priv_decrypt(), returns
without investigating the result pointer on errors, but let's be
conservative here.
Andre Noll [Mon, 26 Dec 2016 15:16:36 +0000 (16:16 +0100)]
build: Remove compatibility check for clock_gettime().
Only Mac OS (which is no longer supported) did not have it. Moreover,
on Linux we used to check whether one needs to link with -lrt to
obtain clock_gettime(), which is necessary for glibc versions 2.16
and earlier (released in 2012). All distributions should have moved
on to newer glibc versions by now, so let's get rid of the check.
Andre Noll [Mon, 18 Jul 2016 18:50:41 +0000 (20:50 +0200)]
Drop support for Mac OS.
This software hasn't been tested on Mac OS for years, so Mac OS
support is most likely already broken.
This commit removes the osx writer and the autoconf tests for core
audio, and adjusts the documentation to not mention Mac OS anymore. It
also effectively reverts commit be1074b4 which introduced an ugly
workaround in interactive.c that was only needed on Mac OS.
Andre Noll [Sat, 31 Dec 2016 20:12:33 +0000 (21:12 +0100)]
Merge branch 'maint'
A trivial conflict in configure.ac and remove/modify conflicts in
web/index.in.html and NEWS. The last file has been renamed to NEWS.md
in master. The conflict was resolved by copying the new text of the
NEWS file from maint (the release notes for v0.4.14) to NEWS.md and
to adjust the formatting to markdown syntax.
Andre Noll [Sat, 31 Dec 2016 15:50:02 +0000 (16:50 +0100)]
Merge branch 'refs/heads/t/simple_error_codes'
Two patches which get rid of the concept of per-subsystem error
codes. The host-compiled error2.c program can be removed, configure.ac
and error.h simplified.
The merge conflicted because both sides modified error.h, but this
was easy to resolve.
* refs/heads/t/simple_error_codes:
Sort errors alphabetically.
Simplify the error subsystem, get rid of error2.[ch].
Andre Noll [Fri, 30 Dec 2016 14:58:41 +0000 (15:58 +0100)]
Merge branch 'refs/heads/t/invalid-ids'
A single patch that was in misc for a while, and two follow up fixups
that were detected after the branch was merged into next.
* refs/heads/t/invalid-ids (cooking for two weeks):
Makefile: Don't compile with -Wformat-signedness unconditionally.
aft.c: Use correct format string for error output.
touch: Refuse to set an invalid image or lyrics ID.
Andre Noll [Wed, 28 Dec 2016 20:40:42 +0000 (21:40 +0100)]
t0004: Specify proper options for ls commands.
The -p option is deprecated, and the default behaviour (if none of -p,
-F and -b is given) will change in v0.6.0. Although the tests succeed
at the moment, it seems prudent to switch to the modern syntax to
make sure the tests won't break when -p is removed and the default
behaviour is changed.
Andre Noll [Wed, 28 Dec 2016 19:38:34 +0000 (20:38 +0100)]
Merge branch 'refs/heads/t/format-signedness'
This series fixes all warnings produced by compiling with
-Wformat-signedness and adds the flag to CFLAGS if the compiler
supports it.
* refs/heads/t/format-signedness (cooking for ~2 weeks):
gcrypt: Fix a few format-signedness issues.
Compile with -Wformat-signedness if possible.
Fix signedness issues in format strings.
Andre Noll [Tue, 27 Dec 2016 18:29:38 +0000 (19:29 +0100)]
resample: Simplify initialization().
In resample_init(), the first check removed in this commit was bogus
because in case there is no parent buffer tree node, we must only
abort if there is no input pending either.
In resample_post_select(), we move up the check of the node status
so that we now call resample_init() only after we know that there
is input available. This makes the second check in resample_init()
pointless as the condition can never be true.
Andre Noll [Wed, 28 Dec 2016 01:08:54 +0000 (02:08 +0100)]
fade: Add documentation for main() and include it in doxygen.
The two mixer implementations for OSS and ALSA are also included,
both of which are only used by para_fade. So it makes sense to include
para_fade as well. main() is the only a non-static function, so let's
doxify that.
Andre Noll [Wed, 28 Dec 2016 12:05:34 +0000 (13:05 +0100)]
Merge branch 'refs/heads/t/wma_fixes'
The series also contains a fix for a silly bug which causes the decoder
to abort on empty output sizes, and a couple of cosmetic cleanups.
* refs/heads/t/wma_fixes (cooking for two weeks):
wmadec: Remove two pointless variables.
wmadec: Remove a pointless cast.
wmadec: Set data size to 0 if nothing was decoded.
wma: Fix packet size calculation.
wmadec: Properly handle empty outputs.
wma_common: Fix typo in log message.
Andre Noll [Tue, 27 Dec 2016 15:29:29 +0000 (16:29 +0100)]
Merge branch 'refs/heads/t/openssl-1.1'
In openssl-1.1 several structures have been made opaque, breaking both
the stream cipher and the public key functions in crypt.c. This series
deals with these issues, trying to minimize the ifdeffery.
* refs/heads/t/openssl-1.1 (cooking for three months):
openssl: RSA fixes for openssl-1.1.
openssl: Use EVP API for AES.
Andre Noll [Mon, 16 Dec 2013 21:18:27 +0000 (22:18 +0100)]
portable_io.h: Provide big-endian versions and use them for aac.
The aac audio format handler code contains some instances that read a
big-endian encoded 32 or 64 bit number from a buffer. While for the
32 bit case there is a helper function aac_read_int32(), the 64 bit
case is open-coded.
We already have similar functions for the conversion of little-endian
entities. This patch adds their big endian counterparts as inline
functions to portable_io.h and changes the callers to use those.
The patch also gets rid of two fprintf() statements in write_portable()
which were commented out for ages.
Andre Noll [Mon, 16 Dec 2013 21:06:44 +0000 (22:06 +0100)]
aac: Switch from unsigned char to char.
The faad library functions take unsigned char pointers while most of
the paraslash code prefers plain char *. It's easier to use char *
in all paraslash functions and cast the arguments of the faad library
function calls than to have a mix of both types.
Andre Noll [Sun, 28 Aug 2016 16:46:22 +0000 (18:46 +0200)]
Simplify the error subsystem, get rid of error2.[ch].
This commit removes error2.c and the surrounding infrastructure of
the build system, getting rid of ~600 LOC.
After the change there are no more subsystems for error codes, and we
don't need to host-compile error2.c any more. Since all executables
now contain the text of every error code, the change has some impact
on the sizes of the (stripped) executables:
Andre Noll [Tue, 20 Dec 2016 14:40:09 +0000 (15:40 +0100)]
Merge branch 'refs/heads/t/base64'
A couple of patches which move the base64 code to a separate file,
and improve on it. Was cooking for several months.
* refs/heads/t/base64:
base64: Speed up decoder by using a table.
base64: Use para_isspace() everywhere.
base64: Trivial whitespace fixes.
base64: Replace Pad64 variable by macro.
base64: Remove unnecessary overflow checks.
base64: Saner semantics for base64_decode() and uudecode().
Move base64 implementation to own file.
Andre Noll [Thu, 8 Dec 2016 23:22:57 +0000 (00:22 +0100)]
wmadec: Remove two pointless variables.
The local variables n and incr of wma_decode_frame() shadow the
values of their counterparts in struct private_wmadec_data, and they
remain constant within the function. Referring directly to the private
structure instead makes the code shorter and improves readability.