[genesis] Missing migration.blob for iota genesis ... command

[lzpap]

Bug description

Generating a genesis with local migration sources produces a wrong starting state without migration data/transactions/objects.

Steps To reproduce the bug

1. Create local obejct snapshot from a recent iota mainnet hornet snapshot:

 ./target/release/iota-genesis-builder --disable-global-snapshot-verification iota --snapshot-path latest-full_snapshot.bin --address-swap-split-map-path swap-split.csv --address-swap-map-path swap.csv --target-network alphanet

where:

• swap-split.csv is:

Origin,Destination,Tokens,TokensTimelocked
iota1qp8h9augeh6tk3uvlxqfapuwv93atv63eqkpru029p6sgvr49eufyz7katr,0x7f5ef659f341e3256d4abd256c66f754fcf78bf574034f5d5ae8d11e55207d20,300000000000,50000000000000
iota1qrkwdhv2juxue3ju8ug9wfyw2q68qh495cjsn87svyfjytka7y3c6nf3lrk,0x5b595105a4b20efe3d7ae28cb0ab907e93023ca21b3463bf55834051a569d04d,300000000000,50000000000000
iota1qrukjnd6jhgwc0ls6dgt574sxuulcsmq5lnzhtv4jmlwkydhe2zvy69t7jj,0x99e4f3ce4bae0763df277b908bd3661e2ba7fc69fb0738b67646e317e0ee98f7,300000000000,0

• swap.csv is

Origin,Destination
iota1qrukjnd6jhgwc0ls6dgt574sxuulcsmq5lnzhtv4jmlwkydhe2zvy69t7jj,0x99e4f3ce4bae0763df277b908bd3661e2ba7fc69fb0738b67646e317e0ee98f7

2. mkdir test_split_config
3. Try creating a genesis via:

./target/release/iota genesis --working-dir test_split_config --num-validators 1 --local-migration-snapshots stardust_object_snapshot.bin --delegator 0x99e4f3ce4bae0763df277b908bd3661e2ba7fc69fb0738b67646e317e0ee98f7

4. The test_split_config directory doesn't have a migration.blob.

Expected behaviour

migration.blob is created and a local network with the migrated state can be started.

./target/release/iota start --network.config test_split_config

Actual behaviour

Missing migration.blob causes the local network to start without migration data.

Solution hint

The following function doesn't check for the emptiness of timelocks_to_split field. Should one add it, migration.blob is created:

iota/crates/iota-genesis-builder/src/stake.rs

Line 47 in cdc9a4c

pub fn is_empty(&self) -> bool {

[miker83z]

The actual problem is in this check:

iota/crates/iota-genesis-builder/src/stake.rs

Line 200 in b772ffa

if !timelock_allocation_objects.to_destroy.is_empty() {

Fixing the is_empty only partially fixes the issue. In !timelock_allocation_objects.to_destroy.is_empty() we check if some timelocks needs to be destroyed and only then we continue creating the allocation. However, if we have just 1 timelock that has an amount enough to cover all the validator delegation, then it is not destroyed but just split! This is the scenario you tested, and in that part of the code it didn't go through creating the correct allocation.
A quick solution would be to just check:

if !timelock_allocation_objects.staked_with_timelock.is_empty()

Because that staked_with_timelock vector is filled in any case a timelock is touched for delegation (split or destroyed).