[Merged by Bors] - ATX V2 Syntactical validation with deps - singular ATXs

[poszu]

Description

Closes #5994

• syntactically validate ATX V2 with deps (previous, commitment and positioning ATXs and poet proofs)
• doesn't support marriage or merged ATXs (denoted in the code with TODO comments)

Test Plan

Added units tests

TODO

• Explain motivation or link existing issue(s)
• Test changes and document test plan
• Update documentation as needed
• Update changelog as needed

[codecov]

Codecov Report

Attention: Patch coverage is 80.41958% with 56 lines in your changes are missing coverage. Please review.

Project coverage is 81.2%. Comparing base (0db7e0a) to head (1ea1d91).

Files	Patch %	Lines
activation/handler_v2.go	81.3%	27 Missing and 17 partials ⚠️
activation/validation.go	50.0%	8 Missing ⚠️
sql/atxs/atxs.go	81.2%	1 Missing and 2 partials ⚠️
activation/handler.go	0.0%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@           Coverage Diff            @@
##           develop   #5949    +/-   ##
========================================
  Coverage     81.2%   81.2%            
========================================
  Files          293     293            
  Lines        31123   31403   +280     
========================================
+ Hits         25276   25507   +231     
- Misses        4198    4229    +31     
- Partials      1649    1667    +18     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[ivan4th]

Querying the full set of ATX fields is not very efficient. For one thing, if we don't need the blob of the commitment ATX here, and doing JOIN on the atx_blobs table (which atxs.Get does) will slow down validation process. Another thing to consider is that multiple separate SQL queries are always slower than a single query, even if the single query fetches some excess fields. Maybe we could consider doing a single query for the deps along the lines of

select epoch, nonce, num_units where id in (?, ?, ?)

using IDs of commitment / prev / positioning ATXs?

[poszu]

For one thing, if we don't need the blob of the commitment ATX here, and doing JOIN on the atx_blobs table (which atxs.Get does) will slow down validation process.

The atxs.Get() doesn't read the blob anymore. It only reads the atxs table.

[fasmat]

I'm not sure a single query is easily doable during the handling of an incoming ATXs, since there are so many branches that need to be considered depending on

• initial vs. non-initial
• merged vs. not-merged
• ATX that resizes post vs. one that doesn't
• etc.

We could consider doing all in a single TX and see if that decreased the validation time.

[ivan4th]

Do we use fields other then those available in the atxs table for previous ATX validation?
That is, do we really need to decode and validate that blob?

[poszu]

We need this because the ATX in the atxs table contains the effective num units. Effective num units are the smallest value of the num units in the ATX and its previous ATX (the ones declared by the ATX itself). If we take a minimum of Atx.NumUnits and PrevATX.EffectiveNumUnits, then no one will be able to grow effectively their PoST size.

[ivan4th]

Interesting why then we're storing atx.NumUnits in the table. Do we have some kind of confusion here?

go-spacemesh/sql/atxs/atxs.go

Line 380 in 1226fcc

stmt.BindInt64(3, int64(atx.NumUnits))

[fasmat]

NumUnits as included by the ATX is only needed to be able to verify the provided post. Effective NumUnits (what we store as NumUnits in the DB) is used to calculate the actual weight of an identity. They are not the same.

To count your NumUnits on disk as effective numunits you need to hold them for at least one full poet round, i.e. you must generate 2 ATXs with that amount of storage - one before the poet round and one after. That's why effective NumUnits is min(previous.NumUnits, atx.NumUnits) (where both are the wire form of the ATX).

[fasmat]

I'm not sure a single query is easily doable during the handling of an incoming ATXs, since there are so many branches that need to be considered depending on

• initial vs. non-initial
• merged vs. not-merged
• ATX that resizes post vs. one that doesn't
• etc.

We could consider doing all in a single TX and see if that decreased the validation time.

[poszu]

bors merge

[spacemesh-bors]

Pull request successfully merged into develop.

Build succeeded:

• ci-status
• systest-status