| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
| |
In some conditions, ln will follow symlinks it isn’t supposed to. Add
the -n switch to keep that from happening.
Closes: https://bugs.gentoo.org/662810
|
| |
|
| |
|
| |
|
|
|
|
|
| |
If badsym can’t be removed, it’ll gum up the works later when trying
to create links, especially since they don’t point anywhere.
|
|
|
|
|
| |
ln now removes an existing symlink when it tries to create one. This
makes it unnecessary to first try to remove the symlink.
|
| |
|
|
|
|
|
|
| |
Any symlink we’re trying to make should only be handled by us. We may
come across a symlink we missed in some previous clean up, but really
should be set properly rather than failing because it exists.
|
|
|
|
|
| |
Look for broken symlinks in the /usr/include/postgresql-${SLOT}
directories. These were made by 2.1. So now we try to clean them up.
|
| |
|
|
|
|
|
|
|
|
|
| |
canoncalise() returns the absolute path a symbolic link points to,
except when the path no longer exists. Adding the -m option – which is
the same for both realpath and readlink – makes it ignore nonexistent
portions of the path.
Bug: https://bugs.gentoo.org/625368
|
| |
|
|\ |
|
|/
|
|
|
|
| |
In such case the user enabled FEATURES="noman" on Portage, there won’t
be a manpage directory to link to. The module now checks that the
directory exists before trying to use it.
|
| |
|
| |
|
|
|
|
| |
do_set() calls do_unset(). No need to call it before do_set().
|
|
|
|
|
| |
The reset action is usless as do_set calls do_unset if active_slot is
something other than (none).
|
|
|
|
| |
Removed unnecessary echo about generating new links.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
No longer use an environment file. All links are in locations that are
searched by default by man and pkgconfig.
Remove several files outright in /etc/eselect/postgresql as they’re
unnecessary.
Added linking for man pages and pkgconfig files.
Rename ${SLOT} to ${slot} as it isn’t a global variable.
Now compatible with the upcoming slot scheme where there is no
dot. (Release 10.0 is just 10.)
|
|
|
|
| |
Among some other cleanups, add pkg-config support.
|
|
|
|
| |
Whoa, can’t believe I let that go for so long.
|
|
|
|
| |
It’s been moved.
|
|
|
|
| |
Decoding those 1 character prefixes is exhausting!
|
| |
|
|
|
|
|
|
|
|
| |
Refactor do_set so that it doesn't create active.links* files. Will need
to refactor unlinker and move its content into do_update.
do_unset now does a lot more work to look for and remove symlinks
managed by this module.
|
|
|
|
|
|
|
|
|
|
|
|
| |
Now that the ebuilds are unified, we don't really need a bunch of files
laying around containing version number and which ebuild is providing
it.
And, pg_config can give us the version number anyway. Sure, we miss out
on revision numbers, but people really shouldn't be using this tool to
get that detailed.
So, given that, rework the do_list.
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Working towards a stateless module, no need to keep track of links or
active slot in a collection of files.
/usr/share/postgresql is symbolic link generated by this module. So, we're
able to determine which real directory it's pointing to, which will be
/usr/share/postgresql-SLOT. A bit of sed magic gets just the slot
portion off the end. Et voila! We no longer need to store the active
slot in a file.
If /usr/share/postgresql doesn't exist, then we haven't set a slot.
|
| |
|
| |
|
|
|
|
| |
Have a new header to create a symlink for. Fixes bug 486674.
|
| |
|
| |
|
|
|
|
| |
get_slots() shouldn't return false positives now.
|
|
|
|
|
|
| |
Fix bug 471866 [1]. Thank you Ben Kohler for reporting and submitting
the patch.
1.) https://bugs.gentoo.org/show_bug.cgi?id=471866
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
| |
module.
Add trailing slash to source path just to be safe.
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
| |
passed was an Emacs regex, which is also a GNU extension of the find utility.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
| |
Unset now calls do_update() as per bug #360467.
do_update() simply calls do_set() rather than do_reset() to avoid loops, and a
reset wasn't really necessary anyway.
Added '-mindepth 1' to 'find' in linker() so that only the contents are returned
and not the directory itself.
Fixed comments regarding linker().
|
| |
|