OpenCl kernel self-test failed when using any version later than v4.2.1 !!

[royce]

I appreciate why this change happened, but from a UX perspective, especially for users who are not developers or very busy and not tracking hashcat release notes, is there any way to at least detect the presence of the bug and explain in a little more detail, something like:

"OpenCL kernel self-test failed - probable AMD OpenCL bug (not hashcat)"

?

[atom]

There's not much to say in terms of error message. This error occurs if the self-test hash was not cracked but it should be cracked. That's all we know.

[hashserious]

There's not much to say in terms of error message. This error occurs if the self-test hash was not cracked but it should be cracked. That's all we know.

If i use --force (cause of specific optimization), does it still self-test?

[intem]

During testing on test versions I use the driver (Win10-64Bit-Radeon-Software-Adrenalin-Edition-18.5.1-May23)
After version (hashcat-5.1.0 + 910) the mechanism broke.

[intem]

So today (hashcat-5.1.0 + 1012) there are no changes in a good direction.

[Rdgeno]

Did you ever find an answer and your right Crimson doesn't work and screws up my system. I was just told if you run skip self test it wont crack the password. I have tried four separate operating systems and three computers Hashcat and Hashcat beta. Windows is the only thing that will use my GPU but I have to disable the test. I might be wrong but if the driver is only broken in hashcat they had to of changed something but I can't find a version other than the latest and the beta that will run on Windows 10 pro with the GPU The-Distribution-Which-Does-Not-Handle-OpenCL-Well (Kali) Unbuntu two versions of Unbuntu and only stinking windows will use the GPU. I hope you respond with good news because I can't even get the beginning of an answer. If they would just say I don't know and Hashcat doesn't work with AMD I could accept that, but they keep referring people to the Hashcat page or blame the driver. If everyone says the sky is blue then the person that says it's red is either color blind a jerk or retarded. The driver is only broken according to Hashcat. I even trick the computer into installing Intels's opencl I guess HP wants you to use theirs so they screw around with  things and you can't load it without a little work. In The-Distribution-Which-Does-Not-Handle-OpenCL-Well (Kali) you can go into the driver and add The-Distribution-Which-Does-Not-Handle-OpenCL-Well (Kali) to the permissions and then it loads but bricks the system because of xorg. Unbuntu same as The-Distribution-Which-Does-Not-Handle-OpenCL-Well (Kali) without the GPU that dam runtime and it doesn't see the GPU. Sorry if I went off a little but point is it's not your fault, I've been at this for almost two weeks now and the best I have is the CPU isn't detected the GPU is but it wont crack hashes because I'm forcing it. Like I said hopefully yo have some good news.

[Szwag1eR]

Can someone provide the latest working version (v5.1.0 +910)? Is there any way to get it? Not only wpa2 cracking doesn't work, hybrid attack function is dead as well

[frnck]

Hi, did you guys fill a bug with AMD?
Could we get more details on the driver bug?

[philsmd]

I would suggest that you at least try to run the beta version from https://hashcat.net/beta . It tries to work around many driver problems. This is the sad part of being dependent on working device drivers, each and every driver version could be different and introduce new problems (but also fix some) and that we generally get no attention/answer from any hardware vendor (atom already reported dozens of easy to reproduce OpenCL driver bugs, the chance of an answer, let alone a fix is very, very, very slim)

[frnck]

thanks will try the beta,
IMO you have a critical mass of users that would be happy to report bugs to AMD if you would funnel them and tell them what to report...