HaxeFoundation
[Cppia] Sys.args contain name of program file
When running haxelib run hxcpp myProgram.cppia -v or bin/Cppia myProgram.cppia -v calling Sys.args () will return ["myProgram.cppia", "-v"] instead of just ["-v"] (like e.g. a cpp compiled version does).
According to https://api.haxe.org/Sys.html argument list should not contain interpreter nor name of program file.
So haxelib run hxcpp <ARGS> just invokes bin/Cppia <ARGS>. bin/Cppia I believe is just the binary produced by building cpp.cppia.Host. Within the host, Sys.args() can be read, but not really written. A patch probably has to be applied to the Sys.args method directly:
https://github.com/HaxeFoundation/haxe/blob/abcf23120811c5efdfd63bf230ed3e4640de08d9/std/cpp/_std/Sys.hx#L54-L57
Perhaps add #if cppia .slice(1) #end plus a method to get the zeroth argument on cppia only? The solution seems a bit messy though :/
I was trying to figure out why formatter benchmark would not run on cppia target. After experimenting and figuring out what happened, I patched formatter locally to remove the first argument (and had to stop formatter from removing the last argument *) and then it ran just fine.
So projects can add a workaround, I was hoping there was something hxcpp could do to make it consistent with C++ target. I mean technically the manual doesn't really mention cppia...
*) The reason why formatter removes the last argument is that when running haxelib run formatter haxelib will add current working directory as a last arg. Luckily haxelib sets HAXELIB_RUN env var, so programs know they run "inside" haxelib and can remove or deal with that extra arg.
Yes, not sure what the solution is here. In nme, I have System.getArgs and System.setArgs which can account for this.
I could hack the arg out, but that would now break your fix.
Maybe I could add host script.cppia -args x y z and hack out everything before '-args'
