The purpose of this PR is to allow including pure headers into multiple translation units without ODR violations.
First and foremost it splits phase2_func_defs into phase2_inline_defs and phase3_regular_defs. Produced .h file gets all declarations (as before), alias definitions and anything template-related, while the rest go to .hpp.
A flag -tu-compatible-h2 is added that stops generated hpp files from including other hpps as it is not necessary to produce a buildable source, but the default behaviour is unchanged. When using the flag it is important to include the resulting .hpp at least somewhere as it is never included automatically.
With a simple-ish cmake functions it makes it possible to produce automatically both header with declarations and inline definitions and a cpp with non-inline definitions. As an example, in my hobby project cmake parses h2s, finds #cppinclude lines and adds them to a new .cpp as regular #include, than includes the generated .hpp.
The purpose of this PR is to allow including pure headers into multiple translation units without ODR violations.
First and foremost it splits
phase2_func_defsintophase2_inline_defsandphase3_regular_defs. Produced .h file gets all declarations (as before), alias definitions and anything template-related, while the rest go to .hpp.A flag
-tu-compatible-h2is added that stops generated hpp files from including other hpps as it is not necessary to produce a buildable source, but the default behaviour is unchanged. When using the flag it is important to include the resulting .hpp at least somewhere as it is never included automatically.With a simple-ish cmake functions it makes it possible to produce automatically both header with declarations and inline definitions and a cpp with non-inline definitions. As an example, in my hobby project cmake parses h2s, finds #cppinclude lines and adds them to a new .cpp as regular #include, than includes the generated .hpp.