Re: [arch-dev-public] [signoff] e2fsprogs-1.41.14-1

2010-12-26 Thread Rémy Oudompheng
signoff x86_64

I have a strange bug when running e2freefrag immediately after an
online resizing. It's the first time I do anything like that, so it
has probably always existed : it seems the number of free blocks is
not updated.

Device: /dev/lvm0/home
Blocksize: 4096 bytes
Total blocks: 1048576
Free blocks: 26816 (2.6%)

Min. free extent: 4 KB
Max. free extent: 129016 KB
Avg. free extent: 2892 KB
Num. free extent: 717

HISTOGRAM OF FREE EXTENT SIZES:
Extent Size Range :  Free extents   Free Blocks  Percent
4K...8K-  :   200   2000.75%
8K...   16K-  :   102   2420.90%
   16K...   32K-  :81   3881.45%
   32K...   64K-  :45   4771.78%
   64K...  128K-  :54  11824.41%
  128K...  256K-  :68  2992   11.16%
  256K...  512K-  :46  4209   15.70%
  512K... 1024K-  :61 10858   40.49%
1M...2M-  :27  9638   35.94%
2M...4M-  : 8  5114   19.07%
4M...8M-  : 6  8028   29.94%
8M...   16M-  : 2  5023   18.73%
   32M...   64M-  : 3 28404  105.92%
   64M...  128M-  :14441956  1648.11%


[arch-dev-public] extra/progsreiserfs duplicate of core/reiserfsprogs?

2010-12-26 Thread Guillaume ALAUX
Hello,

Correct me if I'm wrong but it looks like extra/progsreiserfs
(0.3.0.5-5) is an old duplicate of core/reiserfsprogs (3.6.21-3).

I search for reasons in mailing list archives for having both but found
none.

Nb: Old extra/progsreiserfs PKGBUILD [1] pulls a tarball from
ftp.archlinux.org. Let's remember to remove the tarball if we get rid of
the package.

[1]
http://repos.archlinux.org/wsvn/packages/progsreiserfs/trunk/PKGBUILD
-- 
Guillaume


signature.asc
Description: This is a digitally signed message part


Re: [arch-dev-public] [signoff] cryptsetup 1.2.0-1

2010-12-26 Thread Guillaume ALAUX
On Thu, 2010-12-23 at 23:28 +0100, Thomas Bächler wrote:
 Upstream update, please sign off.
 
 From the announcement:
 
 Changes since version 1.1.3
 
 Important changes
 ~
 
  * Add text version of *FAQ* (Frequently Asked Questions) to distribution.
 
  * Add selection of random/urandom number generator for luksFormat
  (option --use-random and --use-urandom).
 
  (This affects only long term volume key in *luksFormat*,
  not RNG used for salt and AF splitter).
 
   You can also set the default to /dev/random during compilation with
   --enable-dev-random. Compiled-in default is printed in --help output.
 
   Be very careful before changing default to blocking /dev/random use here.
 
  * Fix *luksRemoveKey* to not ask for remaining keyslot passphrase,
  only for removed one.
 
  * No longer support *luksDelKey* (replaced with luksKillSlot).
   * if you want to remove particular passphrase, use *luksKeyRemove*
   * if you want to remove particular keyslot, use *luksKillSlot*
 
  Note that in batch mode *luksKillSlot* allows removing of any keyslot
  without question, in normal mode requires passphrase or keyfile from
  other keyslot.
 
  * *Default alignment* for device (if not overridden by topology info)
  is now (multiple of) *1MiB*.
  This reflects trends in storage technologies and aligns to the same
  defaults for partitions and volume management.
 
  * Allow explicit UUID setting in *luksFormat* and allow change it later
  in *luksUUID* (--uuid parameter).
 
  * All commands using key file now allows limited read from keyfile using
  --keyfile-size and --new-keyfile-size parameters (in bytes).
 
  This change also disallows overloading of --key-size parameter which
  is now exclusively used for key size specification (in bits.)
 
  * *luksFormat* using pre-generated master key now properly allows
  using key file (only passphrase was allowed prior to this update).
 
  * Add --dump-master-key option for *luksDump* to perform volume (master)
  key dump. Note that printed information allows accessing device without
  passphrase so it must be stored encrypted.
 
  This operation is useful for simple Key Escrow function (volume key and
  encryption parameters printed on paper on safe place).
 
  This operation requires passphrase or key file.
 
  * The reload command is no longer supported.
  (Use dmsetup reload instead if needed. There is no real use for this
  function except explicit data corruption:-)
 
  * Cryptsetup now properly checks if underlying device is in use and
  disallows *luksFormat*, *luksOpen* and *create* commands on open
  (e.g. already mapped or mounted) device.
 
  * Option --non-exclusive (already deprecated) is removed.
 
 Libcryptsetup API additions:
 
  * new functions
   * crypt_get_type() - explicit query to crypt device context type
   * crypt_resize() - new resize command using context
   * crypt_keyslot_max() - helper to get number of supported keyslots
   * crypt_get_active_device() - get active device info
   * crypt_set/get_rng_type() - random/urandom RNG setting
   * crypt_set_uuid() - explicit UUID change of existing device
   * crypt_get_device_name() - get underlying device name
 
  * Fix optional password callback handling.
 
  * Allow to activate by internally cached volume key immediately after
  crypt_format() without active slot (for temporary devices with
  on-disk metadata)
 
  * libcryptsetup is binary compatible with 1.1.x release and still
  supports legacy API calls
 
  * cryptsetup binary now uses only new API calls.
 
  * Static compilation of both library (--enable-static) and cryptsetup
  binary (--enable-static-cryptsetup) is now properly implemented by common
  libtool logic.
 
  Prior to this it produced miscompiled dynamic cryptsetup binary with
  statically linked libcryptsetup.
 
  The static binary is compiled as src/cryptsetup.static in parallel
  with dynamic build if requested.
 
 Other changes
 ~
  * Fix default plain password entry from terminal in activate_by_passphrase.
  * Initialize volume key from active device in crypt_init_by_name()
  * Fix cryptsetup binary exit codes.
0 - success, otherwise fail
1 - wrong parameters
2 - no permission
3 - out of memory
4 - wrong device specified
5 - device already exists or device is busy
  * Remove some obsolete info from man page.
  * Add more regression tests for commands.
  * Fix possible double free when handling master key file.
  * Fix pkg-config use in automake scripts.
  * Wipe iteration and salt after luksKillSlot in LUKS header.
  * Rewrite file differ test to C (and fix it to really work).
  * Do not query non-existent device twice (cryptsetup status
 /dev/nonexistent).
  * Check if requested hash is supported before writing LUKS header.
  * Fix problems reported by clang scan-build.
 

Works like a charm.

Signoff x86_64.

-- 
Guillaume


signature.asc
Description: This is a digitally signed message part


[arch-dev-public] glibc kernel version news draft

2010-12-26 Thread Allan McRae

Hi,

I'm going to move the toolchain out of testing, so need a short news 
item about the minimum kernel version.


--draft--

Minimum required kernel version increase

The update to glibc-2.12.2-1 increases the minimum required kernel 
version to 2.6.27.   This kernel version was chosen as it is currently 
the oldest kernel with longterm support upstream.  Users of the Arch 
kernel26 and kerne26-lts packages are unaffected by this change.


Those wanting to continue use an older kernel version will be required 
to rebuild the glibc package with the appropriate adjustment made to the 
--enable-kernel configuration option.


--end draft--

Any comments?

Allan