Martin Kustermann
Martin Kustermann
To me all options on the table seem better then status quo with `envScript` / `envScriptArgs`: * make wrapper shell scripts that populate the environment and call real compiler *...
Thanks for the investigation, @dcharkes ! So it sounds like nobody else is doing what we currently do (i.e. passing along the `cc`/... as well as `envScript` and `envScriptArgs`)? So...
/cc @dcharkes for discussion
> With you having removed the cli_config this reasoning is now broken. I think there were no guarantees that this ever worked with `cli_config` either. A user may have a...
> > So maybe it's not too bad to require users to write dart run -Bmy_package.config.my_key=$PWD/foo.txt > That seems not user friendly, imagine requiring the same thing for commandline arguments...
/cc @natebiggs @osa1
> From talking to @biggs0125, it sounds like there's only one round of tree-shaking, during TFA. If we were to do the lowering in e.g. intrinsics.dart, the guarded code would...
Yes, I agree that from language semantics point of view this is a valid optimization. That also means a compiler should be able to just remove this `D.loadLibrary()` call entirely...
Thanks, @sigmundch - that's also my understanding! I guess my point is the following * The concept of a "deferred library" does not exist and as such one cannot prefetch...
> Could you share this so I can take a look too? Will do. > Is what you are suggesting the equivalent of raising the import-set algorithm to be managed...