some notes before migrating to Nim 1.xx

[jangko]

some reminder to ease the transition to Nim 1.xx

• bncurve cannot compile because of [regression] VM crash when dealing with var param of a proc result nim-lang/Nim#12244. Workaround to make it compile is convert all const to let, but it only increase maintenance cost.
• stint also affected by Nim bug.
  • See nim v1 failed test: test_int_muldiv nim-stint#88
  • Improve compile time test coverage nim-stint#89
• a const in utils/difficulty.nim const bigMin99 = -99.i256 cannot compile because of stint problem no.2. Workaround: write it as const bigMin99 = 0.i256 - 99.i256, but it also weird
• static array return type of modExpInternal in precompiles.nim can be changed to array only and remove the static keyword.
• because of const array indexing zeroes data in object variants at runtime nim-lang/Nim#8015, we cannot change GasCosts table in gas_costs.nim into a const. On the other hand, Nim v1.xx also disallow a macro to access global let anymore. Workaround: we need to introduce a new GasCostKind const table to let it accessible by opTableToCaseStmt compile time proc. It will increase maintenance cost and ugliness, unless we change some design in EVM drastically.
• [regression] Importing more than one module with same name from different packages produce bad codegen nim-lang/Nim#12420
  if the above issues adressed, we are clear to move to Nim 1.xx

[mratsim]

The static keyword in modExpInternal can be changed to {.compileTime.}. Iirc the case where {.compileTime.} triggered Error: request to generate code for .compileTime proc: foo while static worked was fixed. Can't find the PR but related nim-lang/Nim#8051.

For the gas cost table, we may be able to use

let `ResultGasCostsName`*{.compileTime, inject.}: GasCosts = block:

https://github.com/status-im/nimbus/blob/78714295dab4ead5e56c0048c46ae07d8bbed91a/nimbus/vm/interpreter/gas_costs.nim#L358-L379

PR that fuzzed the boundary between compileTime and runtime vars: nim-lang/Nim#12128

[jangko]

nice

[arnetheduck]

[1:54 PM] BOTgitterirc: well regarding #12332, most probably the fix won't be backported

[narimiran]

well regarding #12332, most probably the fix won't be backported

It will be (if no packages break when fixed).

[arnetheduck]

It will be (if no packages break when fixed).

if no packages you are testing break - worth keeping in mind that this is but a small subset, when creating up an upgrade policy for nim.

[jangko]

another nasty bug nim-lang/Nim#12420

[stefantalpalaru]

@mratsim's suggested fixes applied to Nimbus in this branch: https://github.com/status-im/nimbus/commits/nim_v1

Everything compiles and the tests pass on Linux. Let's see the CI.

[jangko]

latest Nim devel branch now can compiles both nimbus and beacon-chain and pass all tests.
closing this issue and we can concentrate to move to Nim 1.0.2