I don't mind removing check from the conditional body itself if that really bothers you, but the error prints and the asserts should stay IMO. The fact that we try to handle the failure from…
I heavily disagree with this one. I started doing it that way and then went with the macros because it was becoming really unreadable. Also, keep in mind that in your example, if you get a user…
While I get the point. I think it's much easier to do an #ifdef BLI_SUBPROCESS_SUPPORT by mistake in that case.
If you use this API you're likely using it from an OS that supports it, so if the…
blender::BlenderSubprocess ? 😕
I would just mention it in a comment.
I thought you meant the is_owner_ property.
That's there because shm_unlink must be called only once.
If you mean the third shm_open parameter, then maybe?
TBH, I have no idea what that is,…
I've made a full pass on all the OS API function calls, so the whole thing is properly checked and more robust.
Yes, that's the whole point. You need to access it from both processes.
The thing is, that parameter must be mutable for some reason.
That's why I'm using data() instead of c_str().
Wide-string might use different number of characters than the multi-byte one
It can have a different number, but never more? Anyway, I copied the Cycles code since I assume it's widely…