[Openmp-dev] [PATCH] [Revisedx2] Initial cmake support

[Hal Finkel]

----- Original Message -----
> From: "Andrey Bokhanko" <andreybokhanko at gmail.com>
> To: "Alp Toker" <alp at nuanti.com>
> Cc: openmp-dev at dcs-maillist2.engr.illinois.edu, "David Chisnall" <dc552 at cam.ac.uk>
> Sent: Monday, June 2, 2014 10:17:50 AM
> Subject: Re: [Openmp-dev] [PATCH] [Revisedx2] Initial cmake support
> 
> 
> 
> 
> Alp,
> 
> With all respect, a few of assertions you made are simply *not true*.
> 
> 
> 
> 
> 
> On Mon, Jun 2, 2014 at 6:02 PM, Alp Toker < alp at nuanti.com > wrote:
> 
> 
> It should be made clear that the current OpenMP runtime CMake build
> system has been in development for some time, including on-list
> discussions in the LLVM community that go back weeks following all
> the best practices we have. The only thing that changed is that C.
> Bergstrom graciously provided the sign-off we needed to integrate
> Jack's work late last week.
> 
> 
> 
> What "discussions... that go back weeks" you are speaking about?!
> 
> Jack started his "On Improving the Build System revisited" thread on
> May 30. This is four days ago, not weeks.
> 
> And since when "all the best practices" include introducing a new
> build system without getting project architect's consent? --
> especially after explicitly asked to do so, a message that you
> conveniently ignored.

To which message are you referring?

> 
> 
> 
> So it's a mischaracterisation to say this happened over the weekend.
> Even if it did that would be on the long side compared to timescales
> seen on llvm-commits.
> 
> 
> What timescales you are speaking about?!
> 
> 
> For reference, we wait for *weeks* for our OpenMP in clang patches to
> be reviewed! And we commit them *only* after explicit consent of one
> of clang code owners -- even if we already got code review from
> someone else.

To be fair, this is a bit different. First, as I've explained, you don't need approval from the code owner, you need approval from some established contributor who is knowledgeable in that part of the code, including design, implementation details and future direction (and, in effect, who is willing to take responsibility for your change). The problem you have with many of the OpenMP patches is that only the code owners are in such a position, but please don't over-generalize. There are many more qualified people for build system changes (although, FWIW, not as many as one might think). Please understand that, as far as we could tell, the future direction (CMake) was clear, and so Alp had all of the knowledge he needed to approve the commit.

> 
> 
> In general it's a good idea to participate in on-list discussions and
> give a heads up if you see people discussing features you have plans
> for. Is there anything else in the pipeline?
> 
> 
> That's *exactly* what we did back in March.
> 
> http://lists.cs.uiuc.edu/pipermail/openmp-dev/2014-March/000055.html

I think the question is more general. Yes, the build system question was posed on the list, but the implementation roadmap was not. The roadmap for the library in general needs to be an open discussion.

 -Hal

> 
> 
> Yours,
> Andrey
> 
> 
> _______________________________________________
> Openmp-dev mailing list
> Openmp-dev at dcs-maillist2.engr.illinois.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/openmp-dev
> 

-- 
Hal Finkel
Assistant Computational Scientist
Leadership Computing Facility
Argonne National Laboratory