Bug 1194494 - 22 Branched installer images consider $releasever to be 'rawhide', not 22, and so install Rawhide packages
Summary: 22 Branched installer images consider $releasever to be 'rawhide', not 22, an...
Keywords:
Status: CLOSED ERRATA
Alias: None
Product: Fedora
Classification: Fedora
Component: anaconda
Version: 22
Hardware: All
OS: Linux
unspecified
urgent
Target Milestone: ---
Assignee: David Shea
QA Contact: Fedora Extras Quality Assurance
URL:
Whiteboard: AcceptedBlocker
Depends On:
Blocks: F22BetaBlocker
TreeView+ depends on / blocked
 
Reported: 2015-02-20 00:29 UTC by Adam Williamson
Modified: 2015-02-23 18:03 UTC (History)
10 users (show)

Fixed In Version: anaconda-22.20.1-1
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed: 2015-02-23 18:03:16 UTC
Type: Bug


Attachments (Terms of Use)

Description Adam Williamson 2015-02-20 00:29:29 UTC
So I'm really not sure whose responsibility this is, but it appears to affect both yum and dnf payloads so filing against anaconda for now.

If you boot a 22 Branched nightly - e.g. http://kojipkgs.fedoraproject.org/mash/branched-20150218/22/x86_64/os/images/boot.iso - and run an install, you get a Rawhide package set with some .fc23 packages, not a 22 Branched package set.

When using the DNF payload there's a log message which shows that it uses this mirror list URL:

https://mirrors.fedoraproject.org/metalink?repo=fedora-rawhide&arch=x86_64

rather than what would be expected:

https://mirrors.fedoraproject.org/metalink?repo=fedora-22&arch=x86_64

with the yum payload there's no handy smoking gun like that, but it certainly seems to do the same thing, as the package set that gets installed has .fc23 packages in it.

This is at least a Beta blocker - "When installing with a dedicated installer image for a specific Fedora flavor, the default package set must be the correct set for that flavor.", https://fedoraproject.org/wiki/Fedora_22_Beta_Release_Criteria#Default_package_set - but we could probably really argue it as an Alpha blocker too. Nominating as Beta for now.

Comment 1 Adam Williamson 2015-02-20 02:03:13 UTC
Ok, I've figured this one out. J'accuse David Shea:

https://github.com/rhinstaller/anaconda/commit/33bcc64bbb0d96c05a4059e6998495c74e6a7e13

The problem with that is this:

[adamw@adam dnf (master)]$ python
Python 2.7.9 (default, Feb  6 2015, 14:42:41) 
[GCC 5.0.0 20150205 (Red Hat 5.0.0-0.7)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> version = '22'
>>> version[:8].isdigit()
True

I think you might want something like:

if len(version) > 7 and version[:8].isdigit():

or something like that.

Comment 2 Adam Williamson 2015-02-20 02:15:50 UTC
changing it to "if len(version) > 7 and version[:8].isdigit():" does work for me in a quick test - at least it 'works' to produce 22 for a 22 Branched nightly. I don't know the precise case where the check should pass and set the releasever to Rawhide, but I can at least test it like this:

>>> version = '22'
>>> len(version) > 7 and version[:8].isdigit()
False
>>> version = '20150214'
>>> len(version) > 7 and version[:8].isdigit()
True
>>> version = '20150214 foobar'
>>> len(version) > 7 and version[:8].isdigit()
True
>>> version = '22-Alpha'
>>> len(version) > 7 and version[:8].isdigit()
False

Comment 3 David Shea 2015-02-20 14:10:26 UTC
Does anyone even use timestamp-based versions anymore?

Comment 4 David Shea 2015-02-20 20:22:54 UTC
(In reply to David Shea from comment #3)
> Does anyone even use timestamp-based versions anymore?

They won't be now!

Comment 5 Dan Mossor [danofsatx] 2015-02-23 18:01:16 UTC
Discussed at today's blocker review meeting [1].

AcceptedBlocker Beta - This bug is a clear violation of the Beta criterion: "When installing with a dedicated installer image for a specific Fedora flavor, the default package set must be the correct set for that flavor."

http://meetbot.fedoraproject.org/fedora-blocker-review/2015-02-23/

Comment 6 Adam Williamson 2015-02-23 18:03:16 UTC
I confirmed this is fixed in 02-21 nightly.


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