IMO this is not really portable, what about mingw on windows?

yes, I was worried about the same thing. mingw on windows doesn't really support concurrent compilation, but it won't fail, the option is just ignored.

Actually mingw does support concurrent compilation, but nproc is not installed and will therefore fail.

-j option is supported, but it has no effect on windows, it's still single threaded (gcc version 8.2.1 mingw)

I've changed that to -j 0 (autodetect), so now the code doesn't rely on nproc

I use gmake quite regularly and used dmake before (it uses -PXX insteal of jXX).
It definitely does compile in parallel, not sure which make flavor you mean that ignores it.
Otherwise I'm fine if 0 means autodetect, but could you please add a more descriptive
title/description to the commit? Always useful to have meaningful git logs! Thanks!

done, I've added a benchmark aswell. On my AMD FX-8300 8C it's 4,6x faster.

Thanks

Surely this is the wrong fix - shouldn't the caller control the fanout by passing make -j 0 if they want this behaviour.

Currently this causes unbounded fan out, which is deeply wrong, and it ignores a caller passed in setting - e.g. make --jobs 1 will be overriden by this change - which is happening when the Rust build attempts to control fanout to the number of CPUs - see compass-rs/sass-rs#53

Compilation on my box now regularly fails with:

STDERR:virtual memory exhausted: Cannot allocate memory
make: *** [src/ast_sel_cmp.o] Error 1
make: *** Waiting for unfinished jobs....
virtual memory exhausted: Cannot allocate memory
virtual memory exhausted: Cannot allocate memory
virtual memory exhausted: Cannot allocate memory
virtual memory exhausted: Cannot allocate memory
virtual memory exhausted: Cannot allocate memory
virtual memory exhausted: Cannot allocate memory
virtual memory exhausted: Cannot allocate memory
make: *** [src/cssize.o] Error 1
make: *** [src/fn_numbers.o] Error 1
make: *** [src/fn_selectors.o] Error 1
make: *** [src/extend.o] Error 1
make: *** [src/inspect.o] Error 1
make: *** [src/remove_placeholders.o] Error 1
make: *** [src/fn_utils.o] Error 1

Sounds reasonable. Open to reverting this change.

ok, sorry, I tested again and you're right. -j 0 is indeed unlimited. The original nprocs limit was removed because it's hard to get the value safely on all platforms. I've reverted the change in #2977

btw optflags https://github.com/sass/libsass/pull/2977/files#diff-b67911656ef5d18c4ae36cb6741b7965L19 also can't be overridden.