Re: [trojita] Build mode, debug symbols, asserts

2015-01-07 Thread Stephan Platz
Hello, On Mittwoch, 7. Januar 2015 18:48:12 CET, Jan Kundrát wrote: Opinions? I vote for combining the user groups to two groups, one with CMAKE_BUILD_TYPE=Debug for developers and debugging and one CMAKE_BUILD_TYPE=RelWithDebInfo or Release. I don't really care whether release build conta

Re: [trojita] Build mode, debug symbols, asserts

2015-01-07 Thread Jan Kundrát
On Wednesday, 7 January 2015 09:25:25 CEST, Pali Rohár wrote: Problem is that -Wall (or -W) is not standard option. Other systems (like Windows) are traditionally using slash instead dash when parsing options. Could be there some problem? In theory, yes. In practice, even MSVC apparently recog

Re: [trojita] Build mode, debug symbols, asserts

2015-01-07 Thread Jan Kundrát
On Tuesday, 6 January 2015 23:25:39 CEST, Thomas Lübking wrote: Occasional -O0 builds on all compilers (to ensure buildability) could maybe done by the CI? What we have now are CI runs for each iteration of each patch. While I do have quite a few CPUs for that, the number is not infinity and t

Re: [trojita] Clang fun with templates

2015-01-07 Thread Jan Kundrát
On Sunday, 4 January 2015 00:58:41 CEST, Thomas Lübking wrote: You may add as many template class Template; as ever you like and wherever you want - clang won't compile, while gcc is happy with the one implementation in class.cpp and no explicit instatiation at all. (Notice that multiple ex

Re: [trojita] Build mode, debug symbols, asserts

2015-01-07 Thread Thomas Lübking
On Mittwoch, 7. Januar 2015 09:25:25 CEST, Pali Rohár wrote: Re debug symbols Should stay by default. Another question is: What do you mean by "default" above? When you "cmake .." w/o further adjustments, this should result in an O2 build. Whether you'd like to support a build type type tha

Re: [trojita] Build mode, debug symbols, asserts

2015-01-07 Thread Pali Rohár
On Tuesday 06 January 2015 23:25:39 Thomas Lübking wrote: > On Dienstag, 6. Januar 2015 18:15:02 CEST, Jan Kundrát wrote: > > - Report all reasonable warnings. > > +1 (for sure) > +1 > > Warnings are IMHO quite useful because they can hint about > > possible problems in the code. > > If there'