Add prevhash to template cache

[Sjors]

This prevents repeating calls to getBlockHeader()

[Sjors]

@ryanofsky is there a way to make methods like getBlockHeader() cache their own result?

[Sjors]

Rebased after #57.

[xyephy]

Hey @Sjors, I think there's an issue with the prev_hash variable in ThreadSv2ClientHandler.

Right now, prev_hash is declared at the top of the while loop but only gets assigned when block_template is null (inside that first if block). This works fine on the first iteration, but on subsequent iterations block_template won't be null anymore (since it gets updated at line 341), so the assignment never happens. This means the comparison if (new_prev_hash != prev_hash) ends up reading uninitialized memory.

I think the fix is pretty straightforward - just move prev_hash outside the loop and update it after each waitNext() call:

uint256 prev_hash;  // Move this before the while loop

while (!m_flag_interrupt_sv2) {
    if (!block_template) {
        // ... create template ...
        prev_hash = block_template->getBlockHeader().hashPrevBlock;
        m_block_template_cache.insert({template_id, std::make_pair(prev_hash, block_template)});
    }

    std::shared_ptr<BlockTemplate> tmpl = block_template->waitNext(options);
    if (tmpl) {
        block_template = tmpl;
        uint256 new_prev_hash{block_template->getBlockHeader().hashPrevBlock};
        
        if (new_prev_hash != prev_hash) {
            future_template = true;
            // ...
        }
        
        prev_hash = new_prev_hash;  // Add this line to update for next iteration
        m_block_template_cache.insert({m_template_id, std::make_pair(new_prev_hash, block_template)});
    }
}

This also has the nice side effect of removing that getBlockHeader() call that was happening every iteration in the original code (the old old_prev_hash line), so it's actually more optimized.

[Sjors]

uint256 is a class, so it's default initialized to 0.

But you're right that it should be set outside the while loop.

[xyephy]

looks good, it can be merged now.

[Sjors]

Unfortunately this was incorrect, see #66.

[Sjors]

Previously old_prev_hash probably should have been m_best_prev_hash.

[xyephy]

Unfortunately this was incorrect, see #66.

let me read through it and have a better understanding of it, sorry about my previous mistake.

[Sjors]

@xyephy I didn't see it either, it's not the best code.