feat(tests): close out eip-7954 checklist - #3546
LouisTsai-Csie wants to merge 4 commits into
Conversation
LouisTsai-Csie
added
C-feat
LouisTsai-Csie
changed the title
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## forks/amsterdam #3546 +/- ##
===================================================
+ Coverage 94.01% 94.44% +0.42%
===================================================
Files 624 624
Lines 36927 36928 +1
Branches 3326 3326
===================================================
+ Hits 34717 34875 +158
+ Misses 1533 1450 -83
+ Partials 677 603 -74
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
The max-initcode gas boundary took its gas limit from `transaction_intrinsic_cost_calculator()`, which returns the greater of the standard intrinsic cost and the EIP-7976 calldata floor. At `MAX_INITCODE_SIZE` the floor is 8,412,608 gas against a standard cost of 556,588, so the case walked the floor's boundary while declaring `INTRINSIC_GAS_TOO_LOW`. EELS maps both rejections to a single error, so it filled clean, but geth's mapper separates them. Take the threshold from `transaction_data_floor_cost_calculator()` directly, expect `INTRINSIC_GAS_BELOW_FLOOR_GAS_COST`, and assert the floor is the greater of the two so that a fork which reverses that cannot move the test onto the other rule unnoticed. Add an `above_floor` case. At the boundary the gas charged equals the gas limit, so a receipt pin there holds under either rule; with slack above the floor the receipt matches only if the floor is what is priced. The deployment itself spends 741,748 gas, so the pinned figure is one that only the floor produces. Rename to `test_max_initcode_size_calldata_floor` and update `test_cases.md` to match.
…uously - `test_max_code_size_deposit_gas`: pin the receipt. The exact fit states the 128 KiB total outright rather than through EIP-8037 grant mechanics, and the short case pins the gas cap, proving the deposit failed as an exceptional halt and not a revert, which post-state alone cannot tell. - Jump tests: pre-set a sentinel in the caller. A stored zero matched both "the CALL returned failure" and "the caller never ran"; now it can only mean the CALL executed and failed. - `test_max_initcode_size_via_create[over_max]`: pin the block access list. The initcode size is rejected before the child address is computed, so the address must be absent and the factory untouched. - `eip_checklist_external_coverage.txt`: state the mechanism behind the `test_coverage` claim (positive sibling one byte smaller, pre-set sentinels) instead of asserting no case is vacuous. - `test_cases.md`: the invalid jump target is a `STOP`, not a `PUSH1`.
|
Nice work! No real comments, just some enhancements to your changes here. I did want to hold off on a merge though bc I investigated a bit an existing test case that a Claude review flagged. Since I ended up investigating and fixing some issues with it, I PR'd it to your branch here at LouisTsai-Csie#19. Nothing else really stands in the way here. I just made a second commit to that PR adding extra pins to some tests since I've been trying to think of ways to make sure our tests don't drift from original intent. Feel free to comment on that PR and continue the conversation! I will happily approve this once we've resolved at least that broken test (first commit in the PR) so lmk what you think of the changes there. |
2cd3d71 to
e2ffc8c
Compare
|
Ignore my push here! I think it essentially didn't do anything. I ended up accidentally rebasing this branch on something and then reset it. Sorry for the trouble 🫣 |
fix(tests): pin the EIP-7954 max-initcode boundary to the calldata floor
|
Created a PR to yours here: LouisTsai-Csie#20 Feel free to add, some tweaks and extra coverage! :) |
3 participants
Description
Fills the remaining coverage gaps, brings the EIP checklist to 100%, hardens assertions that could pass vacuously, and completes the fork transition matrix.
Related Issues or PRs
issue #3217
Checklist
just static<type>(<area>): <title>, where<type>and<area>come from an appropriateC-<type>, respectivelyA-<area>, label. The title should match the target squash commit message.Cute Animal Picture