Commit graph

10 commits

Author SHA1 Message Date
syl20bnr
d822241739 core: use request.el to check elpa archive availability
Add request.el to core/libs
Refactor package.el initialization in configuration-layer.el
Cosmetic improvements to loading messages
Remove redefinition of package-refresh-packages
2015-12-02 23:25:00 -05:00
syl20bnr
629d736b79 core: add package keyword :protected
A protected package cannot be uninstalled nor excluded.
2015-11-19 00:53:57 -05:00
syl20bnr
3dec1ce9ec core: perform local path existence check at configuration time
In order to be able to log the not found path as a warning instead
of crashing.

Also adds unit test to test both valid and invalid local paths.
2015-10-22 22:39:53 -04:00
syl20bnr
7a2e604741 core: remove 'post value for :step property
post value for package :step property is both useless and considered
harmful. It would hijack the determinism of the loading order.

'pre value is used very very rarely and I even think it is not even
used since most of the time a package that must be loaded in pre step
is a bootstrap package. But we keep 'pre since it can be useful to
have it for _exceptional_ cases.
2015-09-24 00:02:02 -04:00
syl20bnr
f294ba3887 core: remove 'private value for :location
While doing the tests I encountered an edge case where a package owned
by the dotfile could be passed either a 'private or 'local location
which is ambiguous (moreover where do we put packages with 'local
location ?).

We remove the ambiguity by removing the 'private value and push the
path ~/.emacs.d/private/local/pkg/ when the owner of a local
package PKG is the dotfile.

When the owner of a local package is a layer then the load path is
in the "local" subdirectory of the layer directory. It adds no value
to use the old 'private location in this case.
2015-09-20 00:49:18 -04:00
justbur
6eab954afe Use + instead of ! for layer categories
Helm seems to treat "!" specially in pattern matching, so having a ! in
the pattern string when traversing directories is problematic. This
change fixes #2737, because as far as I can tell "+" has no special
meaning in a helm pattern.

Of course, we can choose a different character, but I'm fond of "+" as
representing "more layers here".
2015-09-11 00:13:51 -04:00
syl20bnr
6b33031dc5 core: dotspacemacs-distribution, spacemacs-core layer and rename contrib
- Rename contrib directory to layers
- Add new variable dotspacemacs-distribution
- Move spacemacs layer to layers/!distribution
- New layer spacemacs-core in layers/!distribution
- User can now set dotspacemacs-distribution to spacemacs or
  spacemacs-core (default spacemacs)

spacemacs-core is very lightweight layer sufficient to build upon
spacemacs.
2015-09-07 23:44:43 -04:00
syl20bnr
8adc82d1ae core: actually read :disabled-for layer property
Also it is not need to pass a list, example:

```
(setq dotspacemacs-configuration-layers
   '((auto-completion :disabled-for org git)))
```

Will disable auto-completion for org and git layers.
2015-09-03 13:17:02 -04:00
syl20bnr
88d30d7074 core: add :disabled-for keyword for declared layers
Example:

(setq dotspacemacs-configuration-layers
'((auto-completion :disabled-for (org))))

Will not run any org/pre-init-xxx or org/post-init-xxx functions where
xxx is a package owned by the auto-completion layer.

Limitation:

If ownership of an auto-completion package P is stolen by another layer L
then :disabled-for must also be declared with L, otherwise the P will
be configured for org.

(setq dotspacemacs-configuration-layers
'((auto-completion :disabled-for (org))
  (L :disabled-for (org)))

The user should not change the default owner of a layer unless really
required (for instance to use his own fork).
2015-09-02 23:10:30 -04:00
Eivind Fonn
c6ba7d9300 Refactor test system
Should allow easier testing of layers, not just core
2015-08-31 21:16:34 -04:00
Renamed from core/tests/core-configuration-layer-utest.el (Browse further)