Closed Ilia-Kosenkov closed 1 year ago
The ci shows I did not fix yet another usage of cp
, will fix it later today
Ah, probably that's the problem? Sorry, it might be my mistake. I can't remember why I added the lines also to configure.ucrt
. Maybe we can delete the whole file.
Yep, that's the one. I'll test without it as well.
Sorry, I just remembered how it works. On invoke_cargo()
, the compilation is not what's done inside an R package, so it doesn't use Makevars.win
. That's why we need another logic to inject the fake libgcc_eh.a
.
Abandoning in favor of #218
Another attempt to close #209. For successful compilation, linker requires
libgcc_eh.a
. Instead of compiling an empty.c
file into a library file, which is a non-atomic operation, we can resort to simplytouch
inglibgcc_eh.a
, which satisfies the linker's needs. In the end, it gets all of the necessary exports from otherlibgcc*.a
libs, but requireslibgcc_eh.a
to just exist on the search path.