Detect FMT_SPLIT_UNIFIES_CASE error

[loverszhaokai]

No description provided.

[loverszhaokai]

@jfoug

Thanks for your patches for detecting FMT_SPLIT error. Should I change the "PASS" to "FAILED" when
there is FMT_SPLIT warning ?

[magnumripper]

If this had been the normal self-test it should definitely be just a non-fatal warning. But maybe nothing in -test-full should be mere warnings? I'm not sure.

[magnumripper]

We're not counting/summarizing warnings. That is an argument for actually give FAILED for any problem in -test-full. An alternative would be to start counting and summarizing warnings but that may be overkill.

[loverszhaokai]

@magnumripper Thanks. I think I should change it to "FAILED".

[loverszhaokai]

@jfoug @magnumripper

When there is a FMT_SPLIT warning or error, change to "FAILED" by: #1741

[jfoug]

Good, I agree that in -test=full, it should not be a warning, but an error. In that mode, the user is certainly doing an expensive test. I really think however that -test-full probably should be removed from the normal options, and placed into the 'hidden' set.

[magnumripper]

I think you are now failing but still printing "Warning" for the two latter cases? I didn't try it though.

[loverszhaokai]

@magnumripper

I think you are now failing but still printing "Warning" for the two latter cases? I didn't try it though.

Yes.

[...]
Testing: dynamic=md5($p) [256/256 AVX2 8x3]... FAILED (Warning: dynamic=md5($p) should not set FMT_SPLIT_UNIFIES_CASE)
[...]
1 out of 335 tests have FAILED

[magnumripper]

I really think however that -test-full probably should be removed from the normal options, and placed into the 'hidden' set.

I also think we should put most of this in #if !JTR_RELEASE_BUILD or someting like that. Maybe Solar disagrees, and for that matter this is not a lot of code.

[magnumripper]

I think you are now failing but still printing "Warning" for the two latter cases? I didn't try it though.

Yes.

And actually I think for the first case, FAILED will end up printed twice, won't it? I think you need to fix all three. Just give a description of the problem and let the caller add FAILED to it.

[loverszhaokai]

@magnumripper

And actually I think for the first case, FAILED will end up printed twice, won't it? I think you need to fix all three. Just give a description of the problem and let the caller add FAILED to it.

OK.

[magnumripper]

Now that you've got my attention, it's like this:

Benchmarking: dynamic=md5($p) [128/128 AVX 4x3]... FAILED (dynamic=md5($p) should not set FMT_SPLIT_UNIFIES_CASE)

I think the repetition of the format name is redundant and should be dropped. This probably affects most of your fails.

[magnumripper]

I think it should be

Benchmarking: dynamic=md5($p) [128/128 AVX 4x3]... FAILED (should not set FMT_SPLIT_UNIFIES_CASE)

[loverszhaokai]

@magnumripper

I will change it soon. : )

[jfoug]

It's kind of funny. These have been problems for a LONG time, and no one has complained a bit. Toss out a warning, and they get fixed in hours.