Closed sharkcz closed 4 years ago
Thanks for the report. I've temporarily reverted the merge of Roughtime as there are number of issues in the 2.12.0 release which call for a patch release. But I'll make sure to look at this after Roughtime is merged back into master.
Let me know if you need access to a big endian system, I can arrange that.
@sharkcz get me access. This should be in some of the places where I used memcpy, but easier to find with a test system.
I did a blind attempt to fix it. Could you please give a try on #2143?
The PR looks good :-)
roughtime:
roughtime chain ran 14 tests all ok
roughtime nonce ran 4 tests all ok
roughtime server_information ran 14 tests all ok
roughtime_nonce_from_blind:
roughtime nonce_from_blind ran 2 tests in 0.02 msec all ok
roughtime_request:
roughtime request ran 2 tests in 0.01 msec all ok
roughtime_response:
roughtime response ran 23 tests in 4.96 msec all ok
For an account send me your public ssh key to sharkcz at fedoraproject dot org
@sharkcz I updated my PR with ready to merge code. I don't anticipate any issues. I ask you to rerun the PR but I don't really need a test machine.
Still looking good:
roughtime:
roughtime chain ran 14 tests all ok
roughtime nonce ran 4 tests all ok
roughtime server_information ran 14 tests all ok
roughtime_nonce_from_blind:
roughtime nonce_from_blind ran 2 tests in 0.02 msec all ok
roughtime_request:
roughtime request ran 2 tests in 0.02 msec all ok
roughtime_response:
roughtime response ran 23 tests in 5.02 msec all ok
rsa_blinding:
@sharkcz if you have a chance please git it a (last) try. I've made a change to simplify the code and I would sleep better if you tested it.
no problem, I run a rebuild and tests and everything is still OK
@sharkcz since the code to fix this bug generated a coverity warning I have refactored it slightly. Your human CI would be appreciated for branch roughtime_coverity_warning.
@nunojpg, have you removed the branch? Finally I got some time to check it. If it already got merged, then all looks good in our multi-arch CI.
@sharkcz yes, it was merged to master now. Thanks.
After the last batch of updates from Oct 6-7 our CI builds started to fail on ppc64 and s390x arches in the roughtime tests, which has been just added. Both are big endian platforms, so I suspect an endianness issue has been introduced.