Failing DataFormat test on julia 1.6.1

[adigitoleo]

Failing DataFormat test on julia 1.6.1

[jpjones76]

Just a fast update: this test issue has been fixed on branch "main" and I'll be doing a bugfix release including this in a couple of weeks.

[jpjones76]

It looks like your tests didn't try to download the test data set... there should be a check for the test data, with verbose logging of the data download if the data set isn't found. I've looked at this a couple of times over the last two weeks, but I can't understand why that happened to you.

[adigitoleo]

No worries, I'll poke around as time permits and see if I can help. Maybe I can learn something about testing in julia :) Looks like the test is using files from SampleFiles/AH, which contains the following:

shell> ls .julia/packages/SeisIO/JgSIN/test/SampleFiles/AH/
ah1.f  ah2.f  BRV.TSG.DS.lE21.resp  lhz.ah

[adigitoleo]

Curiously, everything is passing just fine on an Ubuntu box:

julia> versioninfo()
Julia Version 1.4.1
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: Intel(R) Xeon(R) CPU E5-1660 v4 @ 3.20GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-8.0.1 (ORCJIT, broadwell)

[adigitoleo]

Do you happen to know of a way to send the output of pkg>test <foo> to a file?

[jpjones76]

One moment. If your tests pass in Ubuntu, then what is the operating system where they're failing?

[adigitoleo]

The test fails on Arch Linux x86_64 with kernel 5.12.9-arch1-1 (julia versioninfo in the OP).

I think it downloads using subversion, which I also have:

❯ svn --version
svn, version 1.14.1 (r1886195)
   compiled May 24 2021, 14:42:10 on x86_64-pc-linux-gnu

Copyright (C) 2021 The Apache Software Foundation.
This software consists of contributions made by many people;
see the NOTICE file for more information.
Subversion is open source software, see http://subversion.apache.org/

The following repository access (RA) modules are available:

* ra_svn : Module for accessing a repository using the svn network protocol.
  - with Cyrus SASL authentication
  - handles 'svn' scheme
* ra_local : Module for accessing a repository on local disk.
  - handles 'file' scheme
* ra_serf : Module for accessing a repository via WebDAV protocol using serf.
  - using serf 1.3.9 (compiled with 1.3.9)
  - handles 'http' scheme
  - handles 'https' scheme

The following authentication credential caches are available:

* Gnome Keyring
* GPG-Agent
* KWallet (KDE)

[adigitoleo]

Maybe it comes from this line:
https://github.com/jpjones76/SeisIO.jl/blob/main/test/DataFormats/test_ah.jl#L54

I was going through the steps in that block individually and for that line I get:

julia>   @test string(u2d(C.misc["ot"]*μs)) == "1990-05-12T04:50:08.7"
Test Failed at REPL[53]:1
  Expression: string(u2d(C.misc["ot"] * μs)) == "1990-05-12T04:50:08.7"
   Evaluated: "1990-05-12T04:50:08.700" == "1990-05-12T04:50:08.7"
ERROR: There was an error during testing

Is there any reason why these are being compared as strings? A similar check is being done 6 lines above by comparing the DateTime values, which works.

[jpjones76]

Wait. Does Arch have a nonstandard implementation of DateTime formatting, like Unix strptime()?

[adigitoleo]

So it looks like the DateTime print format changed in a recent Julia version. I can't quite track down the commit, but it's not an Arch vs Ubuntu thing. I installed Julia 1.6.1 on the Ubuntu machine and I get the same zero-padding behaviour:

julia> versioninfo()
Julia Version 1.6.1
Commit 6aaedecc44 (2021-04-23 05:59 UTC)
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: Intel(R) Xeon(R) CPU E5-1660 v4 @ 3.20GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-11.0.1 (ORCJIT, broadwell)

julia> using Dates

julia> DateTime(Date(2018, 1, 1), Time(8, 15, 42, 10))
2018-01-01T08:15:42.010

Compared to Julia 1.4.1 (default from Ubuntu repos):

julia> versioninfo()
Julia Version 1.4.1
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: Intel(R) Xeon(R) CPU E5-1660 v4 @ 3.20GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-8.0.1 (ORCJIT, broadwell)

julia> using Dates

julia> DateTime(Date(2018, 1, 1), Time(8, 15, 42, 10))
2018-01-01T08:15:42.01

I'll try to find when this was changed, for reference. I suppose it can't hurt to teach myself some git bisect. Your test runners are using 1.5 I think so that narrows it down a bit.

The good news is that without that one line, everything else is passing on 1.6.1 :)

[adigitoleo]

Git bisect was a bit ambitious because I would have to pull older versions of all kinds of dependencies, so I gave up :) But this change was definitely introduced in the 1.6 release, and I think I found the PR:

https://github.com/JuliaLang/julia/pull/35582/files