[Qt-creator] Qt Creator for Linux Kernel Development

[Jason A. Donenfeld]

Hello All,

Responses to each of you are inline below.

Sounds to me like there's a lot of work to be done, still, and I
haven't received any enthusiastic responses from the Qt team of, "yes!
we'd love to work on this and make Qt Creator suitable for kernel
development!" Bummer. Seems like a great opportunity to increase the
tool's user appeal.

On Wed, Nov 23, 2016 at 8:45 AM, Konstantin Tokarev <annulen at yandex.ru> wrote:
> BTW, I'm facing similar issue with QtWebKit: config.h is included into all
> source files, but never in headers, so all #if's in headers are disabled
> and code model does not function properly.

Yep, this is pretty much entirely broken. It's not implemented in the
Qt C++ model, and it straight up does not work in the clang mode
model. The lack of good support for this feature -- project wide
-include file -- is pretty limiting.

On Wed, Nov 23, 2016 at 8:58 AM, Orgad Shaneh <orgads at gmail.com> wrote:
> Which version do you use? Designated initializers support was added in 3.2
> (c56b999ffff249d4cb7dc7e8026a3297b63ff56d).
>
> I now see that the members are not detected correctly (for e.g. Find
> Usages), but parsing looks ok.

I'm using 4.1. No, it still does not work. Here's a screenshot:
https://data.zx2c4.com/designated-init-broken-qt-creator-49b6f51f.png

> You can set PRECOMPILED_HEADER = header.h in your pro file. It should be
> applied to all sources.

Nope, that's certainly not what that field does.

On Wed, Nov 23, 2016 at 10:21 AM, Eike Ziller <Eike.Ziller at qt.io> wrote:
> Are you using the ClangCodeModel plugin? If not, you probably should.
> It does not help for things like find usages and locator, since for this we still use the built-in model,
> and fixes for the built-in model will be very limited.
> But for working with the individual _source_ file it should at least help.
> I can even open Objective-C++ files and get decent highlighting and even
> get some completion for Objective-C ;)

The clang mode model is even more buggy than the built-in one. Not
only that, but it's an order of magnitude slower too, which makes Qt
Creator impossible to use. The only viable usage of the IDE is through
the built-in code model. Hopefully the Qt engineers will put in the
time to bring it up to par with modern times with -include, and add
explicit C support...

> One other issue are header files - there is no way Qt Creator can automatically tell if a
> .h file is supposed to be C, C++, Objective-C, Objective-C++, ....
> One hack workaround for C might be to remove *.h from the text/x-c++hdr mime type in Options > Environment > Mime Types,
> but that would be a big and global hack (if it worked) ;)
> We could probably provide a project setting for a “default” language to use though, which would then also be sharable with all in the project sources with a .shared file.

A project-wide setting would indeed be most welcome for this.


On Wed, Nov 23, 2016 at 5:15 PM, Andrzej Telszewski
<atelszewski at gmail.com> wrote:
> I don't remember exactly, but I used some other feature of QtC (Import
> Project was it?).
> It allowed to easily setup QtC for kernel development.
> It worked pretty well, most of the time.
> Just a quick search:
> http://stackoverflow.com/questions/5417732/howto-prepare-qtcreator-for-linux-driver-kernel-development

The solution listed here is basically wrong and doesn't work. The code
model won't actually parse that #include properly.

>
> I too remember seeing false errors when code model parsed struct members
> initializers.
>
> But this code seems to work just fine now:

Try something slightly more complicated and it totally blows up again.
See the linked screenshot above.

> And yes, although I'm not going to be involved in enhancing QtC, I would
> love to see C receiving decent support.
> QtC is my main (well, the only) IDE, and I write both C and C++ code.

If only the Qt developers cared about this too. Missed opportunity IMHO.

Jason