Bug 1478604 - tzdata from F27 mass rebuild may be bad.
tzdata from F27 mass rebuild may be bad.
Product: Fedora
Classification: Fedora
Component: tzdata (Show other bugs)
Unspecified Unspecified
unspecified Severity unspecified
: ---
: ---
Assigned To: Patsy Franklin
Fedora Extras Quality Assurance
Depends On:
  Show dependency treegraph
Reported: 2017-08-04 20:55 EDT by Matthew Miller
Modified: 2017-08-05 08:51 EDT (History)
2 users (show)

See Also:
Fixed In Version:
Doc Type: If docs needed, set a value
Doc Text:
Story Points: ---
Clone Of:
Last Closed: 2017-08-05 08:51:00 EDT
Type: Bug
Regression: ---
Mount Type: ---
Documentation: ---
Verified Versions:
Category: ---
oVirt Team: ---
RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: ---

Attachments (Terms of Use)

  None (edit)
Description Matthew Miller 2017-08-04 20:55:24 EDT
Description of problem:

The package dateutils fails to build because it's failing part of its standard test suite. That runs

 dzone --next Asia/Singapore 2014-02-22
which should return
 never -> never  Asia/Singapore
but instead gives
 never -> 1901-12-15T37:42:47+08:00     Asia/Singapore
(Note that the Fedora-installed command uses the long name options so it becomes "datezone", but it's built and tested as "dzone".)

Version-Release number of selected component (if applicable):


Additional info:

This is on x86_64.

The test suite is not completely comprehensive, and all other tests succeed; I'm afraid there may be wider problems that just happen to not be hit.
Comment 1 Matthew Miller 2017-08-04 21:07:16 EDT
Oh, I left out something key! It works just fine with tzdata-2017b-1.fc27.

Plus, the F26 binary fails with tzdata-2017b-2.fc27 and works with tzdata-2017b-1.fc27.

This makes me super concerned because the only change between those two releases is the release bump and mass rebuild.
Comment 2 Matthew Miller 2017-08-05 02:13:14 EDT
I guess one major difference is the version of glibc, from which zic comes -- that's what makes these binary databases, right? Looking at the build logs


That seems to have been 2.25.90-1.fc27 (on i686) previously and 2.25.90-28.fc27 (on armv7hl) now. Did something change _intentionally_ in the format?
Comment 3 Matthew Miller 2017-08-05 08:10:50 EDT
Hmmmm. The binary files definitely differ -- they're even different sizes. (410 bytes in -1, and 424 in -2.)

But, comparing the output of `zdump -v` with the two versions gives identical output.
Comment 4 Matthew Miller 2017-08-05 08:22:37 EDT
I thought the difference might be a bug due to the the build host archs, but no, when I rebuild -2 locally in f27 mock, I get the same file. When I build against f26, I get the same file as -1. So, the difference is something in the environment -- possibly a change in zic?
Comment 5 Matthew Miller 2017-08-05 08:51:00 EDT
Ah. Found upstream fix for dateutils


Note You need to log in before you can comment on or make changes to this bug.