summaryrefslogtreecommitdiff
path: root/Build/source/utils/lzma/NEWS
blob: ad85fec1877ca3828c59b434d900a12f99de05f3 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86

LZMA Utils user visible changes
===============================

Version 4.32.6

  * Always use 32-bit integer to hold probability variables. Earlier,
    these were 64-bit on 64-bit architectures, which hurt cache
    efficiency in the CPU, and thus performance of LZMA. 32-bit
    architectures are not affected by this change.

  * Fix a theoretical data corruption bug in the LZMA encoder. It is
    about overflowing a 32-bit integer, whose typical value stays below
    five. I don't know if it is actually possible to construct to a file
    that could make it overflow. Even if it were possible, it would "only"
    make the output file corrupt so that it is 4 GiB too small; there
    are no other security risks. Now the integer is 64-bit to be sure
    it won't overflow.

  * Add support for copying timestamps on operating systems that support
    setting timestamps only by filename, not by file descriptor.

  * Several portability fixes were made.

Version 4.32.5

  * The percentage shown when --verbose is used, works again. Also some
    typos were fixed from the messages printed by --verbose.

  * Several small portability fixes were made.

Version 4.32.4

  * Ignore command line switch --format=alone. This way current scripts
    can be written so that they will produce LZMA_Alone format files
    even with the new command line tool once it is finishes along with
    liblzma.

  * The command line tool now tells if the user tries to decode files
    in the new .lzma format. The message recommends upgrading to newer
    LZMA Utils.

  * Added some internal consistency checks to liblzmadec, so that it
    doesn't crash if given lzmadec_stream whose initialization failed.
    Some applications using zlib and libbzip2 don't check if
    initialization was successful, and expect that error gets caught
    safely later.

Version 4.32.3

  Seems that everything started going wrong once I started calling
  this "stable" instead of "beta".

  * Copy ownership and permissions before closing the output file.
    Fixes <http://qa.mandriva.com/show_bug.cgi?id=35309>.

  * Copy permission bits with some additional sanity checks. It should
    no longer be possible that someone gets access to the output file
    if he/she didn't have access to the original input file. (Maybe
    ACLs could make this more complex still, but currently lzma doesn't
    care about ACLs.)

  * Make "lzma -c foo.lzma" and "lzma -dc foo" work i.e. don't check
    the filename suffix when writing to standard output.

  * Added a copy of GNU GPLv3. It applies only to some files in the
    "tests" directory. The rest of LZMA Utils will stay under the
    older licenses.

Version 4.32.2

  * Doesn't fail with every file anymore.

Version 4.32.1

  * Explicitly check that closing the output file succeeds.
    Fixes a rare but possible silent data loss.
  
  * Close stdout and stderr and check that those operations succeed.
    Again, fixes a rare but possible silent data loss.
  
  * Some help message, documentation, and typo fixes to various places.

  * Added replacement function for GNU getopt_long(). This should make
    the command line tool build on a few more operating systems.