https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105171
--- Comment #23 from Jason A. Donenfeld <jason at zx2c4 dot com> --- The output of crc32_string() is always >0 for that case (I think?). And the case where the user passes an explicit "0" to -frandom-seed was already considered to be "disabled" by the prior code. A 0 seed also plunges that plugin's lfsr into the trivial linear subspace, rendering it inoperable. So all things considered, 0 is a reasonable "off" value for us to use. There might be some room for improvement -- starting with this bug report here -- but things seem fine enough as-is (plus, the bikeshedding in this bug report already doesn't make that road too appealing).
