wrong sha256 checksums? #28

New issue

Open

opened 2026-07-27 16:39:13 -05:00 by scrdcow · 1 comment

scrdcow commented

2026-07-27 16:39:13 -05:00

(Migrated from github.com)

Hi, was downloading cachyos image, Is it the wrong checksums listed? Also tried checking the .img unpacked, and tried the same for ubuntu.

sha256sum ps5-cachyos.img.xz
763d7cc46bcc369228ea7c75162d0d64766b48a668460d4e80476e93a4ddde38 ps5-cachyos.img.xz

Hi, was downloading cachyos image, Is it the wrong checksums listed? Also tried checking the .img unpacked, and tried the same for ubuntu. sha256sum ps5-cachyos.img.xz 763d7cc46bcc369228ea7c75162d0d64766b48a668460d4e80476e93a4ddde38 ps5-cachyos.img.xz

cZdrikjdm commented

2026-08-02 19:11:48 -05:00

(Migrated from github.com)

Still broken on the new build, this looks systemic, not a one-off stale file.

Release: PS5 Linux Image (latest), build 20260731-185014, kernel 7.1.4, patches daa2e49 (published 2026-07-31, i.e. after this issue was opened).

Release page publishes for ps5-cachyos.img.xz: ad25c068fd38cd247fd32fb1be34694a7247d447b1694c129252eb90eb1e8a87
The file actually served hashes to: 84b193c1e9016699ff8d4b1df4d94f155135809187f3975d0886d131747b622d

Evidence it's the published sum, not the download:

Two fully independent downloads — GitHub release asset via browser, and the R2 mirror via curl, are byte-identical, both 84b193c1… (~1.95 GB). A corrupted transfer would differ between sources.
The image itself is healthy: flashed with balenaEtcher (which validates the xz stream during the flash), boots fine, uname -r = 7.1.4 as the release notes say.

Since a completely new build still ships a non-matching checksum, the sums are presumably generated at a different pipeline stage than the asset finally uploaded (e.g. computed before the final artifact/upload, or the asset replaced without regenerating sums).

Still broken on the new build, this looks systemic, not a one-off stale file. Release: PS5 Linux Image (latest), build `20260731-185014`, kernel 7.1.4, patches `daa2e49` (published 2026-07-31, i.e. *after* this issue was opened). - Release page publishes for `ps5-cachyos.img.xz`: `ad25c068fd38cd247fd32fb1be34694a7247d447b1694c129252eb90eb1e8a87` - The file actually served hashes to: `84b193c1e9016699ff8d4b1df4d94f155135809187f3975d0886d131747b622d` Evidence it's the published sum, not the download: - Two fully independent downloads — GitHub release asset via browser, and the R2 mirror via curl, are **byte-identical**, both `84b193c1…` (~1.95 GB). A corrupted transfer would differ between sources. - The image itself is healthy: flashed with balenaEtcher (which validates the xz stream during the flash), boots fine, `uname -r` = `7.1.4` as the release notes say. Since a completely new build still ships a non-matching checksum, the sums are presumably generated at a different pipeline stage than the asset finally uploaded (e.g. computed before the final artifact/upload, or the asset replaced without regenerating sums).