On 14/03/16 19:03, Felipe Magno de Almeida wrote: > On Mon, Mar 14, 2016 at 1:19 PM, Tom Hacohen <t...@osg.samsung.com> wrote: >> On 14/03/16 14:33, Felipe Magno de Almeida wrote: > > [snip] > >>> Or we require GCC-extension, or we just use the uglier version IMO. >>> >> >> Let me just start by saying that clang also supports this GCC extension >> (just to clarify because I think you said it didn't on IRC). > > I've affirmed clang does have it on IRC. Though not very assertive. > >> Anyhow, why is it not valuable enough to test? > > Are you going to double testing time for a path nobody uses currently?
In a similar manner to the old adage: 200% of nothing is still nothing. I'm only going to double the eo suites (unit tests), and that takes nothing, so running more tests there won't take more time. > >> I could literally just >> compile the suite twice once with it once without it. > > Of course we can, are we going to do it? Or else it *will* bit rot. > Besides, slowing calls down by 30% for compilers that don't have this > extension a good compromise for this syntax sugar? Yes we are. It's 30% for eo_add() not "calls". It's not going to be 30% slower, compared to the object creation time it'll probably be more like 1% (guess). > >> That's a strong >> testing plan and it'll be run like this for everyone who has the gcc >> extension, and will only test the fallback for people who don't have the >> gcc extension (which means that both us and jenkins will test both paths >> all the time). > > If it only tests the fallback for people that don't have the gcc > extension that will be 0 persons at this time AFAIK, except for David > Seikel, and I'm sure he won't like when things break only for him or > that his calls cost 30% more. Or maybe he doesn't object? Where did you get that from? I said we test both for everyone that have the gcc extension and only the fallback if you don't have the gcc extension (obvious). > >> I don't see the reason for your objection. > > More costly to test; Slower for "second-class-citizens" that use other > compilers that not clang and gcc; Way more complex; > > I don't see how the benefit of not using the ampersand operator > outweights the costs here, and I write embedded domain specific > languages for C++ for hobby, so I care about syntax. But this time it > is way too costly. As you may have noticed, I previous pushed in &obj, so I'm also "fine" with it. Though I think it's extremely ugly. People objected, I reacted. Try to convince those people. You are probably better off reaching out to them personally because as shown by my push, they don't reply to e-dev, only e-git. Another side comment that is important to make: at the moment we use the gcc extension for gcc compat, and fallback for the rest, but there's nothing preventing us from using C++ lambdas on compilers that support them or whatever, to make it faster on those platforms too. > >> -- >> Tom. > > Regards, > ------------------------------------------------------------------------------ Transform Data into Opportunity. Accelerate data analysis in your applications with Intel Data Analytics Acceleration Library. Click to learn more. http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140 _______________________________________________ enlightenment-devel mailing list enlightenment-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-devel
