that, too, comes across as an exaggeration to me. although there is potential for compilers to implement "reflections on trusting trust", those are the exception rather than the rule. in nearly every case, using any compiler to get things started is enough to get you binaries that correspond to the sources and nothing else. it is nice to be able to verify that there aren't invisible blobs, but such verification is not essential for software freedom, only to show that a certain binary contains a hidden blob (I hope the distinction is clear and meaningful, and not taken as demeaning either one). the verification is made easier and mechanical by having both a free bootstrap path and reproducible builds, but really, starting from two known-independent compiler binaries, using one of them at a time to bootstrap the free compiler that is then used to build the desired program may very well be enough to verify the absence of hidden blobs, provided that the initial compilers are indeed independent, and that the hopefully-free bootstrapped compiler does not borrow library code from them. reproducible builds, of the compiler and of the program compiled by it, would likely help notice in case they do.