afsocket: Remove invalid assertion from afsocket_dd_restore_writer.

[swstephenson]

Asserting that self->writer is null seems to be incorrect.

If the config being initialized is new, self->writer is guaranteed
to be null, however if we are re-initializing an old config after initializing
a new configuration failed, an existing writer may be present if keep-alive
is off (old_config self->writer is set to null if the connection is saved)

Fixes #1722

Signed-off-by: Sam Stephenson sam.stephenson@alliedtelesis.co.nz

[kira-syslogng]

This user does not have permission to start the build. Can one of the admins verify this patch and start the build? (admin please type: ok to test)

[Kokan]

@kira-syslogng ok to test

[kira-syslogng]

Build SUCCESS, the tests were executed on test branch: master and test suite: functions

[bazsi]

The diagnosis is correct, that assert is not true anymore, however we should probably check that cfg_persist_config_fetch() returns NULL in this case, otherwise we would be leaking the self->writer instance.

So exactly one of self->writer or cfg_persist_config_fetch() must be non-NULL but not both.

[kira-syslogng]

Build SUCCESS, the tests were executed on test branch: master and test suite: functions

[kira-syslogng]

Build SUCCESS, the tests were executed on test branch: master and test suite: functions

[kira-syslogng]

Build SUCCESS, the tests were executed on test branch: master and test suite: functions

[lbudai]

LGTM.
Just a note (nothing to do with this in the PR):

• the if statement is a little bit complicated:

 if (!self->writer && item && !_is_protocol_type_changed_during_reload(self, item))

it says that when

• we don't have a writer instance
  • because we set it to NULL when we stored it in the reload cache (aka persist_config) which can happen only when keepalive(yes)
  • OR it is the first time when we try to initialize the writer
• AND we have an item stored in the reload cache (which says that this is not the first time when we try to intialize the writer...)
• AND the protocol is still compatible with the writer,
  then we re-use the writer instance stored in the reload cache...