RetroAchievements scan inconsistent/spotty for PS2 games

Issue #7624 new
retroNUC created an issue

Out of 213 PS2 games that I had on a playlist that were all manually verified as correct RA hashes, only 63 correctly generated hashes when scanned by RetroAchievements. The rest ended up in metadata post-scan as COULDNTFILEHASH.

Unsure if the problem is that these games are CHD format, or whether there’s something wrong with your identification implementation - https://docs.retroachievements.org/Game-Identification/#playstation-2

White trophy icons in image below are my manually set Achievement badges, yellow trophy icons are the new official implementation:

Comments (13)

  1. Christian

    We are using that repo’s release. A hash value of COULDNTFILEHASH means their program returned an error when attempting to hash your file. I have noticed this happens quite a bit more with PS2 CHD files than any other filetype. Unfortunately there isn’t much we can do on our end until the release is updated to correct the issue.

  2. retroNUC reporter

    Hmm. Given that PCSX2 uses the same library (albeit an older version) and manages to recognize these CHDs 100%, that would suggest either:

    • Something broke with PS2 identification between the library release that PCSX2 uses and their latest release, in which case I’d raise an issue with Jamiras on the RA GitHub page
    • Something’s slightly off with the LB library integration

    If you locally roll back to the older library release, that could help narrow it down.

  3. Christian

    Believe you misunderstood what Jamiras was telling you. We aren’t doing any of the logic whatsoever and ARE using the application from the repo he posted. We simply pass your file to THEIR application and apply the result they give when it successfully hashes the file.

  4. retroNUC reporter

    Apologies, deleted my previous post, rahasher does support CHD as part of the scan.

    Looking through the games in my database, I noticed that I had two game entries that were separate (an original and an undub) and two different files on my drive, but both result in the same RA hash because the ELF and partial hash is the same. However, one was picked up correctly in the LB scan, and the other resulted in was COULDNTFILEHASH.

    Manually deleting the RA-related data from the platform XML and redoing the LB scan, both were picked up correctly the second time.

    Perhaps it could be the case you mentioned where if you’re multithreading the hash calls and saturating the CPU, then the timeout for the hash response gets hit on some of these?

  5. retroNUC reporter

    Oh, interesting, I am finding some CHD files that rahasher is failing on!

    When one of the threads does fail, does it kill all other in-progress rahasher threads? Perhaps that’s why we're seeing files that do hash fine come up with COULDNTFILEHASH?

  6. Christian

    There is no communication between processes. I think what is happening is that we have a 1000ms timeout set in app and you have some files that CAN be hashed but take longer than that. We may need to increase that timeout period to assist with that. On my machine failures can take upwards of 5-6s and that just makes scanning this platform unbearable given how often you see it happen.

  7. retroNUC reporter

    An update on this one:

  8. Log in to comment