View Issue Details

IDProjectCategoryView StatusLast Update
0000229fileGeneralpublic2021-02-09 11:26
ReporterJoergS Assigned Tochristos  
PrioritynormalSeverityblockReproducibilityalways
Status assignedResolutionopen 
PlatformVirtualBox imageOSUbuntu (Linux)OS Version16.04
Product Version5.39 
Summary0000229: fit-map-data.testfile does not match fit-map-data.result, thus file-5.39 compilation (Yocto 3.2.1) always fails!
Descriptionwhen building "file-5.39" in a Yocto-3.2.1 environment, its recipe always fails with the following error (excerpt):
| Running test: ../../git/tests/fit-map-data.testfile
| Error: result was
| FIT Map data, unit id 65536, serial 3879446968, Sat May 31 12:00:34 2014, manufacturer 1 (garmin), product 1632, type 4 (Activity)
| expected:
| FIT Map data, unit id 65536, serial 3879446968, Sat May 31 10:00:34 2014, manufacturer 1 (garmin), product 1632, type 4 (Activity)
| ../../git/tests/fit-map-data.testfile: FIT Map data, unit id 65536, serial 3879446968, Sat May 31 12:00:34 2014, manufacturer 1 (garmin), product 1632, type 4 (Activity)

I have tested "file" with the provided fit testdata (git/tests/fit-map-data.testfile) with two other versions of "file" (5.25 and 5.32) and both give the same result as in the Yocto build!
(i.e. NOT the expected result: the timestamp differs for 2 hours!)

please fix the expected result for the fit test.
Steps To Reproducejust execute: (in the file repo)
file tests/fit-map-data.testfile

this will produce:
fit-map-data.testfile: FIT Map data, unit id 65536, serial 3879446968, Sat May 31 12:00:34 2014, manufacturer 1 (garmin), product 1632, type 4 (Activity)

while the expected result for the automated test expects the timestamp to be "Sat May 31 10:00:34 2014" - 2 hours earlier
Additional InformationI don't know-if(/think-that) this is related to the virtual box I am running Ubuntu-16.04 in, I have no other machine to test that;
but the current time (output of the "time" command) is correct...
TagsNo tags attached.

Activities

JoergS

2021-01-15 14:10

reporter   ~0003528

the used commit for the file repo is: 87731415de945660b00f02207d8e9d986ef9b82e
(poky/meta/recipes-devtools/file/file_5.39.bb)

polluks

2021-01-22 14:42

reporter   ~0003529

Exactly two hours? Smells like GMT vs local time conflict...

JoergS

2021-01-25 07:43

reporter   ~0003530

Yeah - some timezone problem in our virtual-machine setup or alike - I meanwhile had the chance to run the "file" testsuite on a native Linux server, and there it worked!

BUT: why is the timezone even considered? shouldn't "file" simply read the absolute timestamp stored in that test-file and just display it?
I mean, if the local timezone is taken into consideration, wouldn't that require many "expected result" files then? one for every existing timezone in the world...

christos

2021-02-05 22:03

manager   ~0003537

The formats use FILE_T_LOCAL so they use localtime instead of gmtime. Perhaps they should not?

christos

2021-02-05 22:04

manager   ~0003538

It is a timezone issue. Changing FILE_T_LOCAL to 0 in softmagic.c should fix it, but is this better (to print times in GMT by default)?

JoergS

2021-02-07 08:40

reporter   ~0003551

Well, that file was generated at the displayed time, but in the timezone of the creator of the file, correct?
So when I look at the output, without any timezone information, it looks as if it was created in my local timezone...

Maybe show the timestamp converted to the local timezone - but in that case, testing would be horrible;
probably the better aproach: also show the timezone of the creator (if that info is available) or the timestamp relative to GMT as you suggested!

christos

2021-02-08 14:52

manager   ~0003552

It appears that the magic for the maps is written on purpose to use leldate (local) instead of ledate (UTC). I think the easiest way is to stick a TZ=UTC before running the tests.

JoergS

2021-02-09 11:26

reporter   ~0003553

export TZ=UTC
before running the test works for me, thank you very much!!!

I think you can close this ticket now, and sorry for even raising this - wasn't a "file" problem at all in the end...

Issue History

Date Modified Username Field Change
2021-01-15 13:58 JoergS New Issue
2021-01-15 14:10 JoergS Note Added: 0003528
2021-01-22 14:42 polluks Note Added: 0003529
2021-01-25 07:43 JoergS Note Added: 0003530
2021-02-05 22:03 christos Note Added: 0003537
2021-02-05 22:04 christos Assigned To => christos
2021-02-05 22:04 christos Status new => assigned
2021-02-05 22:04 christos Status assigned => feedback
2021-02-05 22:04 christos Note Added: 0003538
2021-02-07 08:40 JoergS Note Added: 0003551
2021-02-07 08:40 JoergS Status feedback => assigned
2021-02-08 14:52 christos Note Added: 0003552
2021-02-09 11:26 JoergS Note Added: 0003553