Debian Codegraph @

.cargo/vendor/weezl/Changes.md

raw
1 ## Version 0.1.12
2
3 - Further adjusted a debug assertion for TIFF compatibility. It still had one
4 of its OR conditions formulated incorrectly, relying on equality of the
5 code word mask and the size switch code. In TIFF we hit the limit of the
6 code dictionary (all 12-bit words) one code later than we would detect a
7 code switch that we need to ignore.
8
9 ## Version 0.1.11
10
11 - Adjusted a debug assertion that handled 1-bit code sizes incorrectly in the
12 decoder (relevant for TIFF but non-standard for GIF). In that case the size
13 switch semantics are a bit messy.
14
15 ## Version 0.1.10
16
17 - Reverted changes made in 0.1.9 to the behavior of the decoder under non
18 libtiff-compatibility mode. Trying to read the decoder with an empty output
19 buffer will at least inspect the next symbol and either error or indicate the
20 end-of-stream accordingly.
21
22 ## Version 0.1.9 (yanked)
23
24 - Increased decoding throughput by 3—30% depending on CPU and data.
25 - Added `{encode,decode}::Configuration` as builder types for their respective
26 module. They can be cheaply cloned.
27 - Added `decode::Configuration::with_yield_on_full_buffer` to explicitly opt-in
28 to libtiff compatibility. The decoder will not read or interpret further
29 symbols of the decoding stream when the output buffer is full. This enables a
30 caller to stop fetching symbols and elide an end of stream marker based on
31 out-of-band length information. The decoder might otherwise error, trying to
32 interpret data that does not belong to the stream.
33
34 ## Version 0.1.8
35
36 - Fixed incorrect state after `Decoder::reset`
37 - Added `Debug` to result types
38
39 ## Version 0.1.7
40
41 - Implicit reset is now supported for decoding.
42
43 ## Version 0.1.6
44
45 - Fixed an integer overflow and panic that could occur during decoding.
46 Decoding performance may degrade after long sequences without a reset code.
47
48 ## Version 0.1.5
49
50 - Added `IntoVec` adapters that simplify in-memory de- and encoding. A further
51 'one-shot' interface is exposed in the `Decoder` and `Encoder` themselves
52 which makes the process a one liner in the simplest cases. Contrary to
53 `IntoStream`, these are available in all cases and do not require `std`.
54
55 ## Version 0.1.4
56
57 - Added `IntoAsync` adapters for asynchronous de- and encoding. The interface
58 is implemented only in terms of `futures = 0.3` traits at the moment.
59 - Code sizes smaller than 2 are now allowed for decoding. Since they do not
60 roundtrip it is still an error to use them in the decoder but this avoids
61 accidental panicking, i.e. denial of service, in parsers.
62
63 ## Version 0.1.3
64
65 - Fixes an issue in compression that caused some data to be lost around clear
66 codes. This could corrupt the data stream.
67
68 ## Version 0.1.2
69
70 - Fixes incorrect compression after `Encoder::reset`.
71
72 ## Version 0.1.1
73
74 - The `IntoStream` types now reuse their internal buffers.
75 - Added the methods `set_buffer`, `set_buffer_size` to `IntoStream` for both
76 the encoder and decoder, used to control the automatic allocation.
77 - Deprecated `IntoStream` in configurations without the `std` feature where the
78 type can't even be constructed.
79
80 ## Version 0.1.0 – Aleph
81
82 - Initial major release
83 - Support gif and tiff code size changes
84 - Rough performance numbers:
85 On i5-4690, 8GiB DIMM DDR3 Synchronous 1600 MHz (0,6 ns)
86 ~70MB/s encode, ~230MB/s decode