Re: [Interest] Doing a leanest-possible QtWebEngine build

2016-12-31 Thread Thiago Macieira
Em sábado, 31 de dezembro de 2016, às 17:55:00 BRST, René J. V. Bertin escreveu: > That's all nice and well for optimisation and/or debug flags, but > > - it'd be pretty damn annoying if you can't build a component because the > official optimisation and/or debug flags lead to memory exhaustion o

Re: [Interest] Doing a leanest-possible QtWebEngine build

2016-12-31 Thread René J . V . Bertin
Thiago Macieira wrote: > As you said yourself, something is resetting the flags despite your attempts. > For a Qt module, you're not supposed to do what you're doing. To simplify > everything: don't set CXXFLAGS in your environment when compiling qtbase. That's all nice and well for optimisation

Re: [Interest] Doing a leanest-possible QtWebEngine build

2016-12-30 Thread Thiago Macieira
Em sexta-feira, 30 de dezembro de 2016, às 09:49:11 BRST, Thiago Macieira escreveu: > > > And yet that's the only accepted, official way for qmake. > > > > So how does this work for regular users on a system where they can't > > modify > > the mkspecs files? > > They don't change compiler flags.

Re: [Interest] Doing a leanest-possible QtWebEngine build

2016-12-30 Thread Thiago Macieira
Em sexta-feira, 30 de dezembro de 2016, às 00:59:31 BRST, René J. V. Bertin escreveu: > Thiago Macieira wrote: > > And yet that's the only accepted, official way for qmake. > > So how does this work for regular users on a system where they can't modify > the mkspecs files? They don't change comp

Re: [Interest] Doing a leanest-possible QtWebEngine build

2016-12-29 Thread René J . V . Bertin
Thiago Macieira wrote: > And yet that's the only accepted, official way for qmake. So how does this work for regular users on a system where they can't modify the mkspecs files? >> But above all: the defaults are set in qmodule.pri, also on Linux, with >> >> QMAKE_CFLAGS += -O3 -g >> QMAKE_CXX

Re: [Interest] Doing a leanest-possible QtWebEngine build

2016-12-29 Thread Thiago Macieira
Em quinta-feira, 29 de dezembro de 2016, às 17:40:53 BRST, René J. V. Bertin escreveu: > Thiago Macieira wrote: > > Just edit the mkspec file. That's the official way of changing the > > compiler > > flags. > > Changing a central mkspec file isn't really an option for a build system. And yet tha

Re: [Interest] Doing a leanest-possible QtWebEngine build

2016-12-29 Thread René J . V . Bertin
Ah! Using qmake [...] -after QMAKE_CXXFLAGS-=-O3 QMAKE_CXXFLAGS-=-O2 QMAKE_CXXFLAGS-=-g QMAKE_CXXFLAGS+=-Os begins to have the intended effect, in part. The -O3 -g from qmodule.pri is now replaced with -Os, but something still appends a "-O2 -g" to the CXXFLAGS in the Makefiles. R. ___

Re: [Interest] Doing a leanest-possible QtWebEngine build

2016-12-29 Thread René J . V . Bertin
Thiago Macieira wrote: > Just edit the mkspec file. That's the official way of changing the compiler > flags. Changing a central mkspec file isn't really an option for a build system. But above all: the defaults are set in qmodule.pri, also on Linux, with QMAKE_CFLAGS += -O3 -g QMAKE_CXXFLAGS

Re: [Interest] Doing a leanest-possible QtWebEngine build

2016-12-29 Thread Thiago Macieira
Em quinta-feira, 29 de dezembro de 2016, às 14:42:51 BRST, René J.V. Bertin escreveu: > Is there support for env. variables in whatever .pr* file contains those > local default "-O3 -g" options? No. Just edit the mkspec file. That's the official way of changing the compiler flags. -- Thiago M

Re: [Interest] Doing a leanest-possible QtWebEngine build

2016-12-29 Thread René J . V . Bertin
On Thursday December 29 2016 10:23:05 Allan Sandfeld Jensen wrote: >Os instead of -O2 however. Just invoke qmake in qtwebengine with >WEBENGiNE_CONFIG+=reduce_binary_size (this is also default for embedded >builds). Sadly that doesn't seem to do the trick. At least not when I build against Qt

Re: [Interest] Doing a leanest-possible QtWebEngine build

2016-12-29 Thread Thiago Macieira
Em quinta-feira, 29 de dezembro de 2016, às 12:37:32 BRST, René J. V. Bertin escreveu: > Thiago Macieira wrote: > > What does this have to do with Qt Assistant 5.6.2 text rendering glitch? > > Nothing. I was as surprised as you probably were to see the thread confusion > showing up in GMane. I gu

Re: [Interest] Doing a leanest-possible QtWebEngine build

2016-12-29 Thread René J . V . Bertin
Thiago Macieira wrote: > What does this have to do with Qt Assistant 5.6.2 text rendering glitch? Nothing. I was as surprised as you probably were to see the thread confusion showing up in GMane. I guess the unusual MUA I had to use left in something it didn't show me when I used a reply-to th

Re: [Interest] Doing a leanest-possible QtWebEngine build

2016-12-29 Thread René J . V . Bertin
On Thursday December 29 2016 10:23:05 Allan Sandfeld Jensen wrote: Hi, >QtWebEngine is mostly not built with QMAKE but with Chromium's GYP. And it >doesn't use the QMAKE CFLAGS directly. We have a switch to change GYP to use - Maybe not directly, but the flags used for building QtBase do end up

Re: [Interest] Doing a leanest-possible QtWebEngine build

2016-12-29 Thread Allan Sandfeld Jensen
On Wednesday 28 December 2016, René J.V. Bertin wrote: > Hi, > > In my experience qmake applies the compiler flags that were selected when > QtBase was built. In my case that means they include "-O3 -g". That's fine > for most projects, but not QtWebEngine, where they stress the build host > to th

Re: [Interest] Doing a leanest-possible QtWebEngine build

2016-12-28 Thread Thiago Macieira
Em quarta-feira, 28 de dezembro de 2016, às 20:02:35 BRST, René J.V. Bertin escreveu: > Hi, > > In my experience qmake applies the compiler flags that were selected when > QtBase was built. In my case that means they include "-O3 -g". That's fine > for most projects, but not QtWebEngine, where th