store: Setting the start and end to prior posting changes

[domgreen]

No description provided.

[domgreen]

Not sure if its better to revert to:
return r.loadChunks(ctx, offsets[m:n], seq, offsets[m],offsets[n-1]+maxChunkSize)

or go to:
return r.loadChunks(ctx, offsets[m:n], seq, uint32(p.start), uint32(p.end))

Really need some way of testing this. :/

[domgreen]

I'm still unsure if we should be adding maxChunkSize/maxSeriesSize onto the end parameter when calling loadChunks/loadSeries functions the rng(k) func already adds this ... could be a minor detail.

[bwplotka]

This is it! Will add unit test for this into your branch.

maxChunkSize/maxSeriesSize onto the end parameter when calling loadChunks/loadSeries

I agree - it might be a bug as well, will remove it. We can always add it. Test should catch this hopefully.

[bwplotka]

Added test with naive "partitioner" to simulate multiple parts. This on master triggers:

runtime error: slice bounds out of range - on this branch it works. Awesome 👍

[bwplotka]

And yea, that extra maxSeries/ChunkSize is definitely bug on new version - since we moved to what parts gives us, we should avoid this. Added commit fixing this as well.

[bwplotka]

Thanks for spotting this 👍 Classic golang bug 🤦‍♂️ function closure + loop... 💥

[bwplotka]

Fixes #829

[domgreen]

Thanks for spotting this Classic golang bug function closure + loop...

@devnev spotted it within seconds 😂