fix: avoid panic in IsReverseCharge when reverse-charge flag is absent

[GeoffreySHD]

Description

Keeper.IsReverseCharge in x/tax/keeper/keeper.go performed an unchecked type assertion:

if !ctx.Value(types.ContextKeyTaxReverseCharge).(bool) {

The flag is only stored in the context by the ante handler (custom/auth/ante/fee.go). On any sdk.Context that did not flow through the ante handler — queries, genesis import/export, or calls from other keepers/tools — ctx.Value(...) returns nil, and this assertion panics, taking down the calling process.

Fix

Use the standard comma-ok idiom and default to "no reverse charge" when the flag is absent — which is also the correct fund-flow default (no flag ⇒ tax behaves normally):

reverseCharge, ok := ctx.Value(types.ContextKeyTaxReverseCharge).(bool)
if !ok || !reverseCharge {

Testing

New regression test x/tax/keeper/keeper_test.go (TestIsReverseCharge, 6 subtests):

• absent flag → no panic, returns false (previously panicked — verified the test fails with the panic on pre-fix code)
• wrong-typed flag → no panic, returns false
• explicit true/false → correct result
• event emission behavior unchanged (only on request, only for the no-reverse-charge path)

Verified in a Linux container (golang:1.24, CI parity): go build ./..., go vet ./x/tax/..., full x/tax + x/market test suites, and golangci-lint run (v2.1.6, the pinned version) all pass.

Notes

The same unchecked assertion exists in the Market Module 2.0 branch (#664) — after this lands, that branch can rebase cleanly onto it.

[GeoffreySHD]

Root cause of the red test-ibc-pfm check — not this PR's code:

The failing job died pulling a third-party image during chain setup, before any code from this branch compiled or ran:

ERROR Failed to pull image {"error": "... ghcr.io/strangelove-ventures/heighliner/osmosis/manifests/v25.0.0\": denied"}
failed to create faucet accounts: ... image ghcr.io/strangelove-ventures/heighliner/osmosis:v25.0.0: denied

The strangelove-ventures GitHub org was renamed to amygdala-labs; the old ghcr namespace died with it (repo now resolves to amygdala-labs/heighliner). interchaintest v10.0.1 hardcodes the old path for osmosis in its built-in chain config — every PR would hit this today; the last green run of that workflow was 2026-07-27.

Fix: e718e86 overrides the osmosis image repository per ChainSpec (spec-level Images replaces the built-in entry; Version still applies) to the same v25.0.0 tag under ghcr.io/amygdala-labs/heighliner/osmosis, which is verified pullable. Both PFM tests (TestTerraGaiaOsmoPFM, TestTerraPFM) get the fix.

[GeoffreySHD]

Friendly bump — this is now fully green (31/31 checks, including the real test-ibc-pfm e2e passing with the osmosis image fix aboard). Summary for review: IsReverseCharge did ctx.Value(key).(bool) without the comma-ok check, so any context that skipped the fee/tax ante handler (queries, genesis, external tools) panicked; the fix defaults to false and the PR also repairs the dead strangelove-ventures/heighliner osmosis image pin (org transferred to amygdala-labs) that was failing interchain tests on every PR. Would appreciate a look from @StrathCole @fragwuerdig @hoank101 whenever convenient.

[fragwuerdig]

Thank you. Apart from overly verbosity LGTM,

[GeoffreySHD]

All four comments removed in 54ba2f0 — no code changes.

[fragwuerdig]

DO NOT MERGE YET

Thank you @GeoffreySHD - this will be merged after the Cosmos Labs security disclosure on 28th of September. There will be a rebuild of v4.0.1-patch.3 with official sources published by upstream. This change can be based on that afterwards.

Possibly consensus breaking touching the return behavior of a query into the context values that were setup by the first Ante Handler decorators.

Gonna ship this with a v4.1.0 I think