summaryrefslogtreecommitdiff
path: root/Build/source/utils/lzma-utils/NEWS
diff options
context:
space:
mode:
Diffstat (limited to 'Build/source/utils/lzma-utils/NEWS')
-rw-r--r--Build/source/utils/lzma-utils/NEWS86
1 files changed, 0 insertions, 86 deletions
diff --git a/Build/source/utils/lzma-utils/NEWS b/Build/source/utils/lzma-utils/NEWS
deleted file mode 100644
index ad85fec1877..00000000000
--- a/Build/source/utils/lzma-utils/NEWS
+++ /dev/null
@@ -1,86 +0,0 @@
-
-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.
-