| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
| |
removing all references to the LGPL as a result
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
__dict__ on deallocation.
* Use a fixed-size struct for String (unicode objects are in fact fixed-size)
and store its variant_level that way.
* Don't store Struct, _LongBase, _StrBase variant_level and Struct signature
in a __dict__, but instead have a global dict mapping object IDs to variant
levels, and a global dict mapping Struct IDs to signatures. This is a bit
strange, but easier than correctly freeing the __dict__ (which is stored
at the end of a variable-length struct, so somewhat hard to get at).
* With this change, allocating objects in a loop no longer leaks memory, and
neither does the test case supplied by Luka Renko.
|
| |
|
|
| |
lines in license statement
|
| |
|
|
| |
64-bit archs
|
| |
|
|
|
|
|
|
|
|
|
| |
cases.
This avoids us getting kicked off the bus when trying to put the wrong
number of things in a struct - this used to happen, but was masked by the fact
that the tests ran with service activation, so the service was just killed and
reactivated. Forthcoming changes to get_object make this automatic reactivation
not happen (messages will be directed to the unique name by default, so
stateful communication can work).
|
| |
|
|
| |
attribute are serialized as that object path. Add that attribute to ProxyObject, dbus.Interface and dbus.service.Object.
|
| |
|
|
| |
variants containing byte, not an array of variants containing string
|
| |
|
|
| |
debugging other things, they're too verbose
|
| |
|
|
| |
parentheses around arguments
|
|
|
validation into separate translation units - no *-impl.h remaining
|