Use IEC standard for binary byte units

glib2.0 (2.20.1-0ubuntu3) karmic; urgency=low

  * debian/patches/05_iec_standard.patch: Use IEC standard for binary byte
    units (LP: #369525).

 -- Benjamin Drung <email address hidden> Wed, 29 Apr 2009 23:11:42 +0200

thank you for your bug report, as you can see on the upstream bug that's not something obviously right and we will not apply it in an ubuntu specific way, there has been similar discussion on ubuntu lists

There is a second bug report about it: http://bugzilla.gnome.org/show_bug.cgi?id=309850

It is a bug if an application uses base 10 prefixes when it calculates with base 2. So this is only a bug fix for using the wrong prefixes.

There is currently a discussion on ubuntu-devel-discuss. The users want this bug fix: http://brainstorm.ubuntu.com/idea/17839/

File sizes should be SI units, not IEC units. Disks are measured in SI units for a reason. The disk size and file size in GLib (Nautilus especially) should be SI units, too. Even Apple does this now. http://support.apple.com/kb/TS2419

1,000 byte file = "1.0 kB"
1,000,000 byte file = "1.0 MB"

IEC units are for things like memory that naturally come in multiples of a power of two, to make calculations simpler.

1,024 x 1,024 byte of memory = "1.0 MiB"

Here is an updated debdiff. It follows the upcoming policy: https://wiki.ubuntu.com/UnitsPolicy

glib2.0 (2.23.0-1ubuntu4) lucid; urgency=low

  * Add debian/patches/80_unit.patch to use SI prefixes for base-10 units
    (LP: #369525).

 -- Benjamin Drung <email address hidden> Sat, 19 Dec 2009 17:34:52 +0100

The policy has been approved by the Technical Board now, so I'll make sure it'll land in lucid soon.

Bumping priority since it's a bug now (due to violating the newly instantiated policy).

Note to self: fix the remaining "KB" in the documentation comment as well.

Here is the update debdiff with all KB fixed to kB.

I sent a git-formatted patch against trunk upstream, and updated the test cases to match. I'll apply that updated patch instead (and against 2.23.5). Thanks Benjamin!

This bug was fixed in the package glib2.0 - 2.23.5-1ubuntu2

---------------
glib2.0 (2.23.5-1ubuntu2) lucid; urgency=low

  [ Benjamin Drung ]
  * Add 05_file_size_units.patch: Use SI prefixes for base-10 units. Refer
    to https://wiki.ubuntu.com/UnitsPolicy for more details (LP: #369525).

  [ Martin Pitt ]
  * 05_file_size_units.patch: Fix test cases to match, and replace with git
    formatted patch sent upstream.
 -- Benjamin Drung <email address hidden> Tue, 09 Mar 2010 17:24:26 +0100

This change causes glib to display units differently based on whether an application is running on Ubuntu or on another platform. If a client application wants to integrate smoothly with its environment, it needs to take this into account in its other fields as well, such as preferences dialogs that let users specify a disk cache size or bandwidth speed limit.

I've described some of these at https://bugs.launchpad.net/ubuntu/+source/glib2.0/+bug/538783 and outlined the constraints of patching Transmission here: https://bugs.launchpad.net/ubuntu/+source/transmission/+bug/538504/comments/18

Ideally, this change should be made upstream, so that applications don't have to jump through hoops based on the distro. If upstream and Ubuntu continue to disagree on whether to open boiled eggs from the big or little end, then at least provide a glib mechanism to let application developers know the rules of the road.

Either way, shoving this through during an LTS freeze does more harm than good. This has been a divisive topic for literally years; waiting for 10.10 is not going to hurt anything.

Please revert this change for 10.04.

Attached an argument (screenshot of karmic) to not revert the patch.

glib does use IEC standards now, so closing again. charles' problem is discussed in bug 538783.

Bug 538783 is one issue where this caused trouble, and the patch is heavily debated upstream.

Lucid after beta-1 is not the time for us to spend endless hours on debates like this, so I backed out this patch in glib.

I propose that we just fix nautilus instead. It seems generally safer to not use g_format_size_for_display() at all any more because it's terminally broken (which is what gvfs/system-monitor/palimpsest etc. are doing for that very reason).

Here is the updated 05_file_size_units.patch for g_format_size_for_display() to use IEC prefixes.

Not a release blocker for lucid.

Adding patch-forwarded-upstream since the patch is awaiting review there

This bug was fixed in the package glib2.0 - 2.29.14-0ubuntu1

---------------
glib2.0 (2.29.14-0ubuntu1) oneiric; urgency=low

  * New upstream version:
    * GSettings: don't call g_error() when the schema is missing
    * Use S.I units (lp: #369525)
  * Drop patches included in the new version
  * debian/libglib2.0-0.symbols:
    - updated for the new version
  * 0001-Revert-all-eventfd-related-patches.patch:
    - renamed, don't revert the commits but rather force the eventfd support
      to false in the configure
 -- Sebastien Bacher <email address hidden> Thu, 21 Jul 2011 11:24:19 +0200

The issue is fixed in Oneiric