Difference in behaviour with and without rr

[mgaudet]

Ran into an bug where firefox crashes on startup run under rr, but doesn't without.

The details are in this bugzilla bug; I've waffled a bit about whether or not its worth opening an issue, but I really don't quite understand why rr is changing behaviour.

The associated bug has a pernosco trace that points at the proximate cause of the bug, but it seems like the root might be an rr thing?

[khuey]

rr doesn't support DRI and that's probably confusing the driver.

rr/src/preload/preload_interface.h

Lines 585 to 598 in 90e46d5

	* The background of this hack is that rr doesn't support DRI/DRM
	* currently, so we use the blunt stick of refusing to open this
	* interface file as a way of disabling it entirely. (In addition to
	* tickling xorg.conf, which doesn't entirely do the trick.) It's
	* known how to fix this particular, so let's not let this hack grow
	* too much by piling on.
	*/
	inline static int is_blacklisted_filename(const char* filename) {
	const char* f;
	if (strprefix("/dev/dri/", filename) || streq("/dev/nvidiactl", filename) ||
	streq("/usr/share/alsa/alsa.conf", filename) ||
	streq("/dev/nvidia-uvm", filename)) {
	return 1;
	}