Error during P-1 stage 2 restart: Assertion failed: Stage 2 params mismatch those captured in the .stat logfile!

[tdulcet]

This assert statement does not seem to consider the small prime relocation feature which changes the initial B2_start value:

Mlucas/src/Mlucas.c

Lines 2645 to 2647 in cfc33a8

	// Now compare the params from the restartfile vs those captured in the log:
	if(psmall)
	ASSERT(psmall == i && B2_start == B2/psmall, "Stage 2 params mismatch those captured in the .stat logfile!");

Specifically, it does not seem to account for this section of code which conditionally sets the B2_start value:

Mlucas/src/pm1.c

Lines 1078 to 1086 in cfc33a8

	Small-prime relocation: If B2/psmall >= B1, we map the primes q in [B1,B2/psmall] to [B1*psmall,B2], run
	stage 2 not from B1 to B2 but from B1*psmall to B2, and incorporate them into the stage 2 accumulator when
	each corresponding semiprime q*psmall is hit in the remapped stage 2 interval:
	*/
	pm1_check_bounds(); // This sanity-checks the bounds and sets B2_start = B1 if unset.
	if(B2_start == B1 && B2 >= B1*psmall) { // It's possible user running a S2 with B2/psmall < B1, hence the 2nd clause
	reloc_start = psmall*B1; // Start including relocation-semiprimes once S2 passes this point, and
	B2_start = MIN(reloc_start, B2/psmall); // shift B2_start upward to reflect the fact that primes in [B1,B2/psmall]
	// will be relocated to [B1*psmall,B2].

This prevents restarting stage 2 of some P-1 tests. To fix this, maybe we could change the assert statement to something like:

ASSERT(psmall == i && (B2_start == psmall * B1 || B2_start == B2 / psmall), "Stage 2 params mismatch those captured in the .stat logfile!")

It would also be good if we could get feedback from someone more familiar the P-1 algorithms.

This was first reported by Ken on the forum.