This is merely a historical archive of years 2008-2021, before the migration to mailman3.
A maintained and still updated list archive can be found at https://lists.osmocom.org/hyperkitty/list/OpenBSC@lists.osmocom.org/.
Jacob Erlbeck jerlbeck at sysmocom.deOn 08/31/2013 03:47 PM, Holger Hans Peter Freyther wrote:
> On Fri, Aug 30, 2013 at 06:34:00PM +0200, Jacob Erlbeck wrote:
>
>> Note that in config mode if the tree is searched along the nodes
>> toward the config node and a command is not found this way, a
>> rollback is done by just replacing the vty's node and index member
>> variable by the saved old values which might break the whole thing, when
>> there has been a free() on the way.
> okay, but in this case the config parsing will fail too. So right
> now we should be safe. Thanks for looking into this!
> o
The problem would rather arise in interactive mode when somebody
mistypes a command. It don't know, why outer context searching is
enabled in config mode only and if somebody someday would like to
provide this in non-config sub-nodes, too. Even if not, the scheme used
for UM2K (alloc() in the node command and free() in vty_go_parent()) must
not be used for config nodes. Having the same function (vty_go_parent() )
for just looking up the outer nodes and unwinding the node contexts
effectively forbids side-effects. Perhaps one could add an additional
boolean parameter to select between both modes of operation and
separate the search phase from the unwinding phase.
>
>> @@ -186,36 +130,24 @@ gDEFUN(ournode_exit,
>> gDEFUN(ournode_end,
>> ournode_end_cmd, "end", "End current mode and change to enable mode.")
>> {
>> + if (vty->node >= CONFIG_NODE) {
>
> this would be true for the OM2K node, which is not a configure node.
> Can you elaborate why not using the 'is config node' predicate is not
> necessary?
The 'end' command only makes sense for node >= ENABLE_NODE. If must
not be applied before the vty is in enabled mode, so it's explicitly a no-op
in these cases.
What I'm not sure about is whether the 'end' command should be available
in the non-config nodes, too (or what's the pure cisco way of it). It
doesn't
'end' the configuration then but it returns to the top node in a defined
way,
which might be helpful in automated scripting.
>> def _testConfigNetworkTree(self):
>> self.vty.enable()
>> - self.vty.verify("configure terminal",[''])
>> + self.assertTrue(self.vty.verify("configure terminal",['']))
> where is the change from self.vty.verify to self assertTrue coming
> from? is this related to your change?
>
Not at all, it accidently went into that patch. I'd rather put it into
the PATCH 2/6
'test' patch.