[Qt-creator] GSoC 2018: New feature proposal for Qt Creator

[Orgad Shaneh]

On Mon, Mar 12, 2018 at 11:33 PM, Harri Pasanen <harri at mpaja.com> wrote:

>
>
> > On Mar 12, 2018, at 4:32 PM, Konstantin Tokarev <annulen at yandex.ru>
> wrote:
> >
> >
> >
> > 12.03.2018, 15:46, "Orgad Shaneh" <orgads at gmail.com>:
> >> On Mon, Mar 12, 2018 at 1:38 PM, Sree Gowtham Josyula <
> sreegowthamj at gmail.com> wrote:
> >>> Hi André & Everyone,
> >>>
> >>> Thanks for showing interest in my suggestion. QTCREATORBUG-16246 is
> >>> indeed almost like what I had intended in my previous mail.
> >>>
> >>>> * Would it be enough to have the files on a network share instead of
> rsync'ing them?
> >>>> * Is it really needed to have Clang running on the remote machine?
> Would it be enough to have access to the included headers on remote?
> >>>
> >>> I think Network share you suggest is a good idea. It solves both of
> >>> the above issues. With network sharing, we wouldn't need rsync and we
> >>> wouldn't need to run clang on remote machine.
> >>>
> >>> I will refine my initial proposal and put forth a more detailed
> >>> proposal considering your suggestions and more use-cases asap.
> >>> If you have any other thoughts and suggestions, kindly let me know.
> >>
> >> Hi,
> >>
> >> I strongly suggest not to use network share. We tried that several
> years ago (with SMB), and it was awful. Parsing takes forever over the
> network. Working locally and using rsync before build works much better
> (once you have ssh keys set up).
> >>
> >> We have a local partial copy of the sysroot, which includes the include
> directory, and the shared libraries that are linked with our application
> (for each platform we support).
> >
> > Why not to go further and get full copy and toolchain locally?
> >
>
> Network share would in a proper setup be the local disk of the compilation
> server.
>
> Also, one cannot copy locally the toolchain if the server is different OS,
> unless a cross compilation environment is setup, which in itself is a major
> piece of work.
>

That's right. What we did was to create dummy compilers that return the
same values as the real compiler when called with -v and -dM etc. This is
no longer needed, as a "custom compiler" can be created instead.

I suggest to read the values from the remote compiler, and create a local
custom compiler with the same values (except include directories which
should be adapted to the local sysroot location).

- Orgad
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.qt-project.org/pipermail/qt-creator/attachments/20180313/dd8bf4bf/attachment.html>