[Gambas-devel] Cygwin patch

[Benoît Minisini]

Le 12/09/2016 à 21:09, Tobias Boege a écrit :
> On Mon, 12 Sep 2016, Bastian Germann wrote:
>> Hi!
>>
>> I saw someone included my patch in 3.9.0. Thanks for that. Here are
>> two more Cygwin patches.
>> Now I am able to build every package except the gb-v4l (which is
>> deprecated anyway) and gb-inotify package, which is dependent on Linux
>> kernel and required for the IDE unfortunately. I am currently working
>> on a cygport.
>>
>> To address the inotify issue: Can I just patch the requirement
>> definitions and calls to that package out of the IDE or does it
>> implement a feature that is absolutely necessary? Then I would use
>> fswatch [0] or Qt's QFileSystemWatcher to implement an alternative
>> component. Actually in the Todo of gb-inotify there are two
>> alternative (outdated?) libs listed that the author wants to have a
>> look at. How does the FreeBSD port handle the file watching?
>>
>> What are your thoughts on this?
>>
>
> I'm the author and I myself don't really plan to extend the component in
> the foreseeable future. Portable file system monitoring would come in the
> form of a new component anyway. The TODO at this time is only a reminder
> where to start should I have some excess time in the future.
>
> I didn't even know the IDE used gb.inotify. There is, AFAICS, no code in
> the IDE that uses gb.inotify. I think you can just remove the component
> dependency from the .project file and be done with it. I suspect the
> reason it is added by default is that the FileView and DirView classes
> in gb.form, which are used in the IDE, detect whether gb.inotify is loaded
> and if so, behave more responsively by using the component.
>
> If that's the whole reason the issue might even be fixed in the main
> repository by removing the explicit dependency and adding a
>
>   Try Component.Load("gb.inotify")
>
> or something to the start procedure of the IDE. That would work in all
> cases.
>
> Regards,
> Tobi
>

I confirm that 'gb.inotify' is optional and, of course, I should have 
done what Tobias suggests instead of explicitely check the component.

Some people still do not use Linux. :-)

-- 
Benoît Minisini