translators-packages team mailing list archive

[Radon Rosborough <2120743@xxxxxxxxxxxxxxxxxx>]

Public bug reported:

I've been looking into why some of my Ubuntu systems are taking longer
than they should to perform APT upgrades. One reason is the locale-gen
step is compiling 17(!!) additional locales on top of the one that's
actually needed, which is en_US.UTF-8. This makes locale-gen take 50
seconds rather than 2-3 seconds.

When I run dpkg-reconfigure locales, and configure only the en_US.UTF-8
locale to be enabled, I would expect (as a user) for only that locale to
be enabled. However, this is not the case. The other locales are enabled
regardless. It is further confusing to look at /etc/locale.gen and see
that the additional locales are commented out there per the dpkg-
reconfigure session, yet are used anyway.

The first issue I found is that the language-pack-en-base package, which
is installed by default and seems to be required as a dependency of
GNOME in order to provide translation files for the en_US.UTF-8 locale,
is installing a file /var/lib/locales/supported.d/en that overwrites
/etc/locale.gen by adding additional locales.

As far as I can tell, the only way to work around that is to add a file
to /etc/dpkg/dpkg.cfg.d with something like "path-exclude
/var/lib/locales/supported.d/en", and reinstall the package. Maybe
#152945 would eliminate the need to do this; I'm not sure.

But then I discovered a second problem, even worse, which is that
reinstalling language-pack-en-base (which may also occur automatically
during any system upgrade) will forcibly update the contents of
/etc/locale.gen to uncomment all the lines of the additional locales!
This is because the package postinst script is invoking
/usr/share/locales/install-language-pack, which is invoking locale-gen
with an undocumented(?!) positional argument that activates an
undocumented(?!) feature of having it write to locale.gen, rather than
reading from it.

At this point, I am kind of stuck on figuring out how do I configure my
system so that only en_US.UTF-8 is made available as a locale, without
the changes being automatically reverted by APT. Is this behavior
intended? Is it documented anywhere? Am I missing some way to opt out of
the additional locales being added back every time?

This behavior occurs for me on all my systems where I am running Ubuntu
24.04 with GNOME. I believe all versions of language-pack-en-base are
affected but the current one is 1:24.04+20250130.

** Affects: language-pack-en-base (Ubuntu)
     Importance: Undecided
         Status: New

-- 
You received this bug notification because you are a member of
Translators Packages, which is subscribed to language-pack-en-base in
Ubuntu.
https://bugs.launchpad.net/bugs/2120743

Title:
  postinst script overwrites /etc/locale.gen

Status in language-pack-en-base package in Ubuntu:
  New

Bug description:
  I've been looking into why some of my Ubuntu systems are taking longer
  than they should to perform APT upgrades. One reason is the locale-gen
  step is compiling 17(!!) additional locales on top of the one that's
  actually needed, which is en_US.UTF-8. This makes locale-gen take 50
  seconds rather than 2-3 seconds.

  When I run dpkg-reconfigure locales, and configure only the
  en_US.UTF-8 locale to be enabled, I would expect (as a user) for only
  that locale to be enabled. However, this is not the case. The other
  locales are enabled regardless. It is further confusing to look at
  /etc/locale.gen and see that the additional locales are commented out
  there per the dpkg-reconfigure session, yet are used anyway.

  The first issue I found is that the language-pack-en-base package,
  which is installed by default and seems to be required as a dependency
  of GNOME in order to provide translation files for the en_US.UTF-8
  locale, is installing a file /var/lib/locales/supported.d/en that
  overwrites /etc/locale.gen by adding additional locales.

  As far as I can tell, the only way to work around that is to add a
  file to /etc/dpkg/dpkg.cfg.d with something like "path-exclude
  /var/lib/locales/supported.d/en", and reinstall the package. Maybe
  #152945 would eliminate the need to do this; I'm not sure.

  But then I discovered a second problem, even worse, which is that
  reinstalling language-pack-en-base (which may also occur automatically
  during any system upgrade) will forcibly update the contents of
  /etc/locale.gen to uncomment all the lines of the additional locales!
  This is because the package postinst script is invoking
  /usr/share/locales/install-language-pack, which is invoking locale-gen
  with an undocumented(?!) positional argument that activates an
  undocumented(?!) feature of having it write to locale.gen, rather than
  reading from it.

  At this point, I am kind of stuck on figuring out how do I configure
  my system so that only en_US.UTF-8 is made available as a locale,
  without the changes being automatically reverted by APT. Is this
  behavior intended? Is it documented anywhere? Am I missing some way to
  opt out of the additional locales being added back every time?

  This behavior occurs for me on all my systems where I am running
  Ubuntu 24.04 with GNOME. I believe all versions of language-pack-en-
  base are affected but the current one is 1:24.04+20250130.

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/language-pack-en-base/+bug/2120743/+subscriptions