Make template classes use NS_LOG_* macros

Make template classes use NS_LOG_* macros

[Stefano Avallone, 2017-03-27 15:58:22]

For a description, see:

Bug 2668 - NS_LOG can not be used in templates

[Peter Barnes, 2017-06-01 21:05:26]

I'd like to hid the details in two macros,

NS_LOG_TEMPLATE_DECLARE
NS_LOG_TEMPLATE_DEFINE(component)

This will reduce implementation errors, and hides the anomalous appearance of
'g_log' among the member variables.

Can we switch from & to *?  (see comment in log.h)

This needs documentation, in log.h and manual.

https://codereview.appspot.com/320790043/diff/1/src/core/model/log.h
File src/core/model/log.h (right):

https://codereview.appspot.com/320790043/diff/1/src/core/model/log.h#newcode418
src/core/model/log.h:418: LogComponent & GetLogComponent (const std::string
name);
I can't remember why I suggested using a reference rather than a pointer.  As
used in classes

    LogComponent & g_log;

it looks weird.  I'd prefer 

    LogComponent * g_log;

https://codereview.appspot.com/320790043/diff/1/src/network/utils/drop-tail-q...
File src/network/utils/drop-tail-queue.h (right):

https://codereview.appspot.com/320790043/diff/1/src/network/utils/drop-tail-q...
src/network/utils/drop-tail-queue.h:62: LogComponent & g_log;     //!<
redefinition of the log component
I'd prefer to hide this in a macro in log.h:

#define NS_LOG_TEMPLATE_DECLARE  LogComponent & g_log

https://codereview.appspot.com/320790043/diff/1/src/network/utils/drop-tail-q...
src/network/utils/drop-tail-queue.h:85: g_log (GetLogComponent
("DropTailQueue"))
I'd prefer to wrap this in a macro in log.h:

#define NS_LOG_TEMPLATE_DEFINE(component)  g_log (GetLogComponent (component))

https://codereview.appspot.com/320790043/diff/1/src/network/utils/net-device-...
File src/network/utils/net-device-queue-interface.h (right):

https://codereview.appspot.com/320790043/diff/1/src/network/utils/net-device-...
src/network/utils/net-device-queue-interface.h:361: static LogComponent & g_log
= GetLogComponent ("NetDeviceQueueInterface");
Now I'm confused.  Why isn't this once in the c'tor?  Instead of in every
function which wants to log?

[Stefano Avallone, 2017-06-05 22:30:55]

https://codereview.appspot.com/320790043/diff/1/src/core/model/log.h
File src/core/model/log.h (right):

https://codereview.appspot.com/320790043/diff/1/src/core/model/log.h#newcode418
src/core/model/log.h:418: LogComponent & GetLogComponent (const std::string
name);
NS_LOG_* macros call g_log.IsEnabled, hence g_log cannot be a pointer.

https://codereview.appspot.com/320790043/diff/1/src/network/utils/drop-tail-q...
File src/network/utils/drop-tail-queue.h (right):

https://codereview.appspot.com/320790043/diff/1/src/network/utils/drop-tail-q...
src/network/utils/drop-tail-queue.h:62: LogComponent & g_log;     //!<
redefinition of the log component
On 2017/06/01 21:05:26, Peter Barnes wrote:
> I'd prefer to hide this in a macro in log.h:
> 
> #define NS_LOG_TEMPLATE_DECLARE  LogComponent & g_log

Done.

https://codereview.appspot.com/320790043/diff/1/src/network/utils/drop-tail-q...
src/network/utils/drop-tail-queue.h:85: g_log (GetLogComponent
("DropTailQueue"))
On 2017/06/01 21:05:26, Peter Barnes wrote:
> I'd prefer to wrap this in a macro in log.h:
> 
> #define NS_LOG_TEMPLATE_DEFINE(component)  g_log (GetLogComponent (component))

Done.

https://codereview.appspot.com/320790043/diff/1/src/network/utils/net-device-...
File src/network/utils/net-device-queue-interface.h (right):

https://codereview.appspot.com/320790043/diff/1/src/network/utils/net-device-...
src/network/utils/net-device-queue-interface.h:361: static LogComponent & g_log
= GetLogComponent ("NetDeviceQueueInterface");
It's not in the constructor because these functions are static, hence they
cannot access the member variables of object instances.

I defined a new macro to be used within static templeate methods

[Peter Barnes, 2017-06-05 23:03:27]

Very nice.  Ok with me to push.

Note the latest patch has extraneous stuff from wifi, I think from your latest
patch r12921	39d9ce52e1af

Peter

[Stefano Avallone, 2017-06-06 09:22:43]

On 2017/06/05 23:03:27, Peter Barnes wrote:
> Very nice.  Ok with me to push.

Pushed with changeset 12922:d5736db31887, thanks.

Please note that python bindings need to be re-scanned, but python bindings
generation is broken (at least for me) after applying the patch proposed to fix
bug #2658.
 
> Note the latest patch has extraneous stuff from wifi, I think from your latest
> patch r12921	39d9ce52e1af

Yes, sorry, my Mac doesn't like the upload.py script...

Stefano