I assume the fix is to rewrite it to va_start(ap, (int) option)? It would take a very sadistic compiler to turn the current code into actual UB though, IMO.

Out of curiosity, with what clang version is that? 3.9.0?

@bnoordhuis Thanks for the reply.

va_start(ap, (int) option) isn't a correct fix, IMO. The promotion rule is still in va_start. I don't know whether there is a fix without changing uv_loop_option interface.

Can you explain when it would make a material difference? sizeof(uv_loop_option) == sizeof(int) and they're both integral types so I'm having a hard time seeing how default argument promotion could turn bad. Clang's warning borders on language lawyering, IMO.

I don't know whether there is a fix without changing uv_loop_option interface.

That would break source compatibility so it's not an option for the v1.x branch but it could be considered for the v2.x branch.

sizeof(uv_loop_option) == sizeof(int)

uv_loop_option is an enum and the compiler can choose to use a short (or even a char) as internal representation, so the above assumption can be wrong. For instance, even with gcc, you will likely break the build by enabling -fshort-enums.

That particular warning seems to have been taken out of clang again and we don't build with -fshort-enums nor are there compilers in our support list where sizeof(uv_loop_option) == sizeof(int) doesn't hold.

No action appears necessary at this time so I'm going to close this out. Please leave a comment or open a pull request if this affects you.