[Development] A better headersclean test

Thiago Macieira thiago.macieira at intel.com
Thu Dec 18 18:32:57 CET 2014


On Thursday 18 December 2014 13:53:38 Giuseppe D'Angelo wrote:
> Il 18/12/2014 04:57, Thiago Macieira ha scritto:
> > Drawback:
> > It increases the build time considerably (25 to 33%) due to having a lot
> > more files to compile, and always without PCH.
> 
> Are those files part of the tests? I.e. -developer-build -nomake tests
> should not have any performance drawback?

No, that's not how it's implemented. The headers are checked when you make the 
module in question.

Right now, this is enabled by passing -headersclean to configure. If the ML 
agrees, it will be enabled by -developer-build too (but then you can disable 
with -no-headersclean).

I've considered replacing mkspecs/features/qt_headersclean.prf with an 
automatic build of each header, which would imply qt_module.prf would somehow 
need to write out its list of headers (it's actually already done by syncqt) 
and the name of the module to a global file. I think it's possible, but it has 
two drawbacks: 1) it still requires new modules to remember to add the test, 
which 16 out of current 25 modules forgot, and 2) people would still forget to 
run the unit test.

-- 
Thiago Macieira - thiago.macieira (AT) intel.com
  Software Architect - Intel Open Source Technology Center




More information about the Development mailing list