| 1 |
On Thu, 2020-09-03 at 15:06 +0200, Florian Schmaus wrote:
|
| 2 |
> CAUTION: This email originated from outside of the organization. Do not click links or open attachments unless you recognize the sender and know the content is safe. |
| 3 |
> |
| 4 |
> |
| 5 |
> Portage's env-update currently transforms the environment information |
| 6 |
> from /etc/env.d into /etc/profile.env, which is typically sourced by |
| 7 |
> every user session, setting up its environment. |
| 8 |
> |
| 9 |
> However, /etc/profile.env is not sourced by systemd user |
| 10 |
> services. Instead, for the definition of a systemd user session |
| 11 |
> environment, the 'environment.d' machinery exists. Unfortunately, up |
| 12 |
> to now, env-update does not produce a profile.env equivalent for this |
| 13 |
> machinery, causing issues for systemd user services. For example, |
| 14 |
> an emacs daemon run a user systemd service does not have a complete |
| 15 |
> PATH (bug #704412 [1]), because some PATH components are injected by |
| 16 |
> packages via /etc/env.d. For example, an LLVM ebuild may set |
| 17 |
> PATH="/usr/lib/llvm/9/bin". |
| 18 |
> |
| 19 |
> This commit changes env-update so that, after profile.env has was |
| 20 |
> generated, a systemd user session environment configuration file named |
| 21 |
> |
| 22 |
> /etc/environment.d/gentoo-profile-env.conf |
| 23 |
|
| 24 |
Are files /etc/environment.d/ ordered in anyway? If so it might be a good idea
|
| 25 |
to name the file 10-gentoo-profile-env.conf or similar.
|
| 26 |
|
| 27 |
Jocke |
Archives