Hi,
I’m working on a solution, basically moving the plugin.c rule in an intermediate Makefile.am.inc in plugins/epan, plugins/wiretap, plugins/codecs.
When it works I’ll push it as a separate change.
Thanks,
Jaap
> On 19 Jan 2018, at 12:10, João Valverde <joao.valverde@xxxxxxxxxxxxxxxxxx> wrote:
>
> Right, well done on the analysis. The make-plugin-reg.py script wants to be passed "plugin_wtap" for wiretap.
>
> On 19-01-2018 08:02, Jaap Keuter wrote:
>> On 19-01-18 04:46, João Valverde wrote:
>>> If the plugin is loaded I don't see how the build system could have any
>>> influence either.
>> Hi João,
>> That was what puzzled met too...
>> I started to dig down further and found that indeed the plugin was loaded
>> (obviously) and they are called to register their registration function, as
>> provided by the generated plugin.c.
>> #1 is the plugin.c file generated in the autotools build, and
>> #2 is the plugin.c file generated in the cmake build
>> Looking at #1 versus #2 shows that #1 has an empty plugin_register() function,
>> while #2 has the correct wtap_register_plugin() call.
>> plugin.c generation is controlled by the build system, so that is the cause of
>> the difference between the two. Generation of #1 comes from a generic
>> Makefile.am.inc, which is geared to generating 'plugin' (being dissector
>> plugin!), while the CMakeLists.txt file calls for generating a 'plugin_wtap'.
>> So, the generation of #1 for the other kinds of plugin (!dissector) has to be
>> reconsidered.
>> Thanks,
>> Jaap