feat: retain sky 0.6.17 4250FF66, 4319D3A2 and D967386B helper profiles - #39
Open
suixincrazy wants to merge 2 commits into
Open
Conversation
Desktop 26.818.4152.0 and 26.818.5229.0 both ship @oai/sky 0.6.17-202608171537-pr-1300023-7efba775c041, but each carries a distinct codex-computer-use.exe. All three 0.6.17 binaries seen so far (29D5E113, DB8F4486, 4250FF66, 4319D3A2) are exactly 1,895,728 bytes with a byte-identical package.json, so the sky version alone cannot select a profile and the strict OriginalSha256 gate rejects the new builds outright. Both new binaries were proven to be region-level clones of the DB8F4486 baseline before adding a profile: a byte diff against the known-good helper shows the only differences below the highest patch region (0x0014B128 + 8) are the PE checksum bytes at 0x158-0x159, and all five patch regions read back identical. The five region offsets are therefore unchanged from the other 0.6.17 profiles. Patched hashes were computed from the real patched artifacts and both profiles pass test-computer-use-helper-win10-patch.ps1. On 26.818.5229.0 the end-to-end screenshot regression against an actively animated Win32 window returns 28 unique JPEG frames with no thread or handle growth on the helper process and no SetIsBorderRequired / E_NOINTERFACE fallbacks.
…818.8289 Desktop 26.818.8289.0 ships a fifth distinct codex-computer-use.exe under the same @oai/sky version string 0.6.17-202608171537-pr-1300023-7efba775c041. All five binaries are 1,895,728 bytes with a byte-identical package.json, so the version string cannot select a profile; the full SHA-256 has to. Clone safety was established before reusing the validated byte patterns: a whole-file diff against the known-good 4319D3A2 baseline shows the only difference below the highest patch region is the PE checksum at 0x158-0x159, and all five patch regions compare identical. The rip-relative references inside the mta-worker-wrapper blob were additionally confirmed to resolve to the same imports (CreateThread, CloseHandle, RoInitialize, RoUninitialize) and the same call target, and both section tables are byte-identical. Patch offsets are unchanged from the other 0.6.16/0.6.17 profiles.
suixincrazy
changed the title
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Desktop keeps shipping new
codex-computer-use.exebinaries under the same@oai/skyversion string0.6.17-202608171537-pr-1300023-7efba775c041. All four0.6.17 binaries seen so far are byte-identical in every patch region, so the Win10
screenshot patch applies unchanged; only the hash pair differs.
This PR retains three of them:
0.6.17-4250FF660.6.17-4319D3A20.6.17-D967386BEach was cleared before adding by a whole-file byte diff against an already
validated baseline: below the highest patch region (
0x0014B128+8) the onlydifference is the PE checksum at
0x158-0x159, and all five patch regionscompare
identical=True. For D967386B the import references inside the175-byte
mta-worker-wrapperblob were additionally resolved and shown totarget the same imports (
CreateThread,CloseHandle,RoInitialize,RoUninitialize) and the samecall rel32destination, with byte-identicalsection tables.
Verification:
test-computer-use-helper-win10-patch.ps1passes for all three new labels(offline apply + patched-hash match + rollback to the original hash)
helper hash matches
PatchedSha256, and a real animated-window regressionproduced 28/28 unique JPEG frames with zero
SetIsBorderRequired/E_NOINTERFACE/0x80004002occurrencesNote on the earlier commit message in this branch:
62cc4ffsays "All three0.6.17 binaries" while listing four hashes. It should read "All four"; the
pushed commit is left as-is and this description carries the correction. With
D967386B the count is five.