| Commit message (Collapse) | Author | Age | Files | Lines |
... | |
|
|
|
|
|
|
|
|
|
|
|
| |
It is expected that users may have the same key set multiple times if
they use separate files in an 'override' type of situation. In this
case the higher-numbered file wins and we should silently ignore the key
set in the lower-numbered file.
This silences the warning message that we previously issued in that
case.
https://bugzilla.gnome.org/show_bug.cgi?id=664288
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This patch is a substantial cleanup of the 'dconf update' command. The
biggest 'clean up' aspect is less dependence on catch-all 'throws'
clauses and better error messages for specific failures.
The cleanup also brings a new feature: keyfiles are now read in a
pre-defined order so that keys defined in higher-numbered files have
precedence over low-numbered files (ie: if the same key is in a file
'10_a' and '20_b' then '20_b' will "win"). This portion of the patch is
based on a patch from Josselin Mouette.
https://bugzilla.gnome.org/show_bug.cgi?id=670494
|
| |
|
|
|
|
| |
https://bugzilla.gnome.org/show_bug.cgi?id=663547
|
|
|
|
|
|
| |
It is now possible to have arrays of nullable types and some places
where we had the equivalent non-null type are now considered to be
errors.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Fix a crash caused by attempting to dereference index -1 of a string.
This is very obviously a bug, but has only become a problem on OpenBSD
where the allocator will apparently hand out addresses at the very start
of a page with invalid memory immediately before.
Problem caught by Antoine Jacoutot.
https://bugzilla.gnome.org/show_bug.cgi?id=662271
|
|
|
|
| |
These changes are upstream by now.
|
|
|
|
|
|
|
|
| |
We were incorrectly checking for NULL as a return value from Dir.open()
when that function as bound as throwing an exception on error -- so
catch the exception instead.
https://bugzilla.gnome.org/show_bug.cgi?id=662158
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The dconf service can not presently be run at the system level and it
doesn't make sense to support locks on user-level databases. It also
became clear that most distributors and sysadmins would rather work
directly with text files anyway, so we supported that directly instead.
For this reason, 'set lock' support has never been properly implemented.
All the plumbing was added for it though, which means we have it
appearing on the API of the client library and documented in the help of
the commandline tool. This is misleading, since these functions do
nothing at all (and actually contain bugs anyway since their
do-nothingness was never actually tested).
For now, we rip out these functions. We can add them back later if we
decide to support this properly.
|
|
|
|
|
| |
These should be in the same format as the system database keyfiles and
fairly similar to the GSettings keyfile backend format.
|
|
|
|
| |
Fixup from last commit.
|
|
|
|
|
|
|
| |
The bash completion file is meant to be sourced and not executed
directly.
https://bugzilla.gnome.org/show_bug.cgi?id=651936
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
| |
- regenerate the file when the lock dir timestamp changes
- send the proper dbus message
- write an actual value for the lock in the GVDB
(otherwise the reader won't see it)
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
| |
Make the dconf command-line tool slightly more friendly by adding a help
command and invoking that on an unknown command instead of aborting.
|
|
|
|
| |
Fixes build with gcc 4.6.
|
| |
|
|
|
|
| |
I meant to check for the file not existing, which is ENOENT.
|
|
|
|
| |
This reverts commit 66dcd38beb93f8c2051b5318508515e396b0e5dd.
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
Rename set_lock to set_locked again (and same SetLock -> SetLocked).
Add missing gtk-doc bits, clean up some that are no longer there.
Bump gtk-doc dependency.
|
|
|
|
|
|
|
|
|
|
|
| |
You need to type 'make clean' before trying to rebuild dconf after this
update (in fact, 'git clean' may be appropriate to clean out all the
extra files that libtool leaves around).
It's quite likely that this completely breaks dconf builds on non-ELF
platforms (and maybe some ELF ones). Please report any problems to
bugzilla. I'm very happy to accept patches to fix portability (via
./configure-time checks).
|
|
|
|
|
|
|
|
|
|
|
|
| |
- the API is now completely implemented
- update 'engine' API to be vala-bindable without annotations
- update GSettings backend to new engine API
- drop the readtype non-sense
- build/api fixups for editor and commandline tool
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This utility scans /etc/dconf/db/ looking for directories named *.d.
For each directory name.d that it finds, it reads each file contained
within as a keyfile containing a list of keys to populate a dconf
database with.
That database is then written to 'name' (ie: the directory name with the
'.d' removed), the old file is invalidated, and a signal is sent over
the system DBus to indicate the change.
This tool facilitates updating of dconf databases by sysadmins who
prefer to follow the 'drop a text file here' approach to system
management.
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
| |
- move to new GSettingsBackend API
- add 'length' parameter, update callers
|
|
|
|
| |
fix a silly bug that the testing found
|
| |
|
| |
|
|
|
|
|
| |
- add a sync set() call to client API
- GCancellable/GErrorify some APIs
|
|
Only very preliminary functionality for all of these.
|