fix(vendor): 6b/6c review corrections — pre-send guards, accumulating staging, trade-note exemption, drag-over tab switch, full-stack sells
Some checks are pending
Headless portability / portable-headless (ubuntu-latest) (push) Waiting to run
Headless portability / portable-headless (windows-latest) (push) Waiting to run
Headless portability / linux-graphical (push) Waiting to run
Headless portability / linux-vulkan (push) Waiting to run
Some checks are pending
Headless portability / portable-headless (ubuntu-latest) (push) Waiting to run
Headless portability / portable-headless (windows-latest) (push) Waiting to run
Headless portability / linux-graphical (push) Waiting to run
Headless portability / linux-vulkan (push) Waiting to run
All thirteen findings, each anchored in recovered bytes or pc reads:
Buy All now runs retail's four PRE-SEND guards in order (pyreal and
alt-currency affordability, container and item slot capacity; strings
recovered from .rdata at 0x007b57b4/0x007b5750) — a rejected batch can
no longer destroy the staged list. Staged adds ACCUMULATE with the
5000 cap ("I can't possibly sell you that much!..." @0x007b59d8) and
the shop rows decrement/restore per RemoveFromShop. The max-value sell
rejection exempts trade notes — the raw bytes at 0x005d1add are `not`
(bitwise), not the pseudo-C's misleading `!`, and the early ret skips
the min check too. BF_RETAINED gates selling end to end (the bit was
already on ClientObject; AP-164's three claims were all false once
traced — RETIRED). Dragging over the vendor window auto-opens the
Selling tab per UpdateDragOver — with a correction to the review's own
citation: token 0x100000cd is the SELLING page, the guard is
"don't reopen the current tab." Sells are full-stack-only (three
retail sites; "Cannot sell part of a stack" @0x007b57ec) and Sell Item
acts on the global selection unconditionally. The confirm string gains
its byte-true trailing '?', dies with the session, staged-row
highlights repaint, dead guids unstage with retail's shopping-list
notice, and move-to-use no longer walks to targets the dispatch would
refuse.
AP-162 narrowed, AP-164 retired, AP-167/AP-168 filed honest.
Clean-room complete solution: 11,508 passed / 4 skipped / 0 failed.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
This commit is contained in:
parent
92ea3977b6
commit
c68ad1e646
11 changed files with 1314 additions and 99 deletions
|
|
@ -411,6 +411,31 @@ public sealed class SelectionInteractionControllerTests
|
|||
Assert.Empty(h.Transport.Uses);
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// F11 (Slice 6b/6c review): a FAR target the eligibility gate will
|
||||
/// refuse (not owned by the player, not useable) must never kick off a
|
||||
/// speculative local approach at all — the prior <c>RequestUse</c>
|
||||
/// ordering called <c>BeginApproach</c> before the eligibility check,
|
||||
/// so an out-of-range unusable target still walked the player toward
|
||||
/// it even though the dispatch immediately below was always going to
|
||||
/// reject. The existing <see cref="RejectedWorldUseReleasesItsBusyReservation"/>
|
||||
/// test never configured an approach at all (<c>TryGetApproach</c>
|
||||
/// returns false unconditionally), so it could not have caught this —
|
||||
/// this test explicitly combines "far" with "rejected."
|
||||
/// </summary>
|
||||
[Fact]
|
||||
public void FarUnusableTargetIsRejectedWithoutApproaching()
|
||||
{
|
||||
var h = new Harness();
|
||||
h.Query.Useable = false;
|
||||
h.SetApproach(closeRange: false);
|
||||
|
||||
h.Controller.SendUse(Target);
|
||||
|
||||
Assert.Empty(h.Movement.Approaches);
|
||||
Assert.Empty(h.Transport.Uses);
|
||||
}
|
||||
|
||||
[Fact]
|
||||
public void SynchronousMovementCallbackCannotDuplicateWorldUse()
|
||||
{
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue