Test main for failure seen in #7378 for Y2038 bug#7399
Test main for failure seen in #7378 for Y2038 bug#7399Carreau wants to merge 4 commits intouutils:mainfrom
Conversation
Carreau
force-pushed
the
touch-y2038
branch
3 times, most recently
from
71e2c73 to
8efc9ab
Compare
|
GNU testsuite comparison: |
|
GNU testsuite comparison: |
|
Those three failures seem like actual failures / y2038 bugs ? Which I'm going to guess ends up in the |
|
I don't understand why this change is in a separated PR from #7378 |
|
#7378 test (and fix) compatibility with gnu touch, i.e. that
So really #7378, test the mapping from 2 character years to 4 character years, and could just test This PR tests that dates just before, at, or just after y2038 bug (even when already a 4 character years) get actually written on filesystem with the correct timestamp (they are not on 32 bits systems). So it tests for a potential bug, independently of whether the year is 2 or 4 characters. |
|
Ok, that makes sense, thanks ! |
|
GNU testsuite comparison: |
|
I guess one question I don't know how to answer is what should touch do on 32 bit systems, and is it a bug in filetime or something else that should tell you (or prevent you) from setting a wrong timestamp. |
Well, that may be an unpopular opinion, but I'd say that's where our dedication to sticking to GNU's behavior should end. |
|
I tend to agree, but maybe an error (instead of a silent failure) as long as this project is tested on 32 bits is still worth it ? |
tests/by-util/test_touch.rs
Outdated
| seconds + deltas | ||
| ); | ||
| let file = format!("Y2038_{}_{}", date, deltas); | ||
| ucmd.args(&["-t", &date, &file]).succeeds().no_output(); |
There was a problem hiding this comment.
Maybe it would be simpler/more concise to just specify the time argument directly? So the three test cases would be
ucmd.args(&["-t", "203801190314.06", "f1"])
ucmd.args(&["-t", "203801190314.07", "f2"])
ucmd.args(&["-t", "203801190314.08", "f3"])
There was a problem hiding this comment.
I expanded to 3 tests one just before, one at and one just after.
|
@Carreau Can you solve the conflict ? |
|
GNU testsuite comparison: |
|
GNU testsuite comparison: |
|
gentle nudge |
|
GNU testsuite comparison: |
|
some jobs are failing |
Yes, because the goal of the PR here is to add tests for a behavior that is already failing and was not tested. I don't think I know how you want to fix it (if you want to fix it, or even if fixable), or if you want me to ignore on some platforms/aarch. (rebased) |
|
GNU testsuite comparison: |
|
GNU testsuite comparison: |
|
GNU testsuite comparison: |
It's unclear to me how the failure seen in uutils#7378 is related to the change in the PR, it just happen that the test need to try touching in 2068. The 2^32 error in the test makes me strongly think of the year 2038 bug, so I added a test for that.
|
It's been 6+ month, I rebased, but next time I think aboutthis PR, I think I may just close it. Feel free to reuse the code/takeover. |
|
GNU testsuite comparison: |
4 participants
It's unclear to me how the failure seen in #7378
is related to the change in the PR, it just happen that the test need to try touching in 2068.
The 2^32 error in the test makes me strongly think of the year 2038 bug, so I added a test for that.
This will obviously fail on 32bits platforms, I don't know how you want to handle failure.