Proposed as a Blocker for 29-final by Fedora user jlanda using the blocker tracking app because:

 server dvd continues warning about prerelease.

Pre-release notices criteria: No notices or alerts about pre-release status may be shown as part of installation or critical path actions on release-blocking desktops.

server netinst is affected too

Affected isos so far:

server-dvd, server-netinstall, everything-netinstall, wks-netinstall

Unaffected ones:
wks-live

Affected isos are affected by https://bugzilla.redhat.com/show_bug.cgi?id=1642089 too, so I suspect that actually it's the same bug.

I believe this is just the nightlies compose configuration and RC configuration that releng uses for spinning RCs disables the prerelease warning.

I guess you're right Kalev. However, I can confirm that Workstation Live 20181022 doesn't contain the warning while other images do. Also, Workstation Live 20181014 does contain the warning! So something fishing is going on here. I believe releng (and possibly) need to look into this.

CC Mohan.

+1 blocker technically, whether it means a code change or just a releng configuration change

I would rather believe that the RC candidate should have this warning switched off, as nightly builds are technically still Beta, ergo a pre-release. 

I am +1 blocker to make sure this gets a proper treatment.

From #fedora-releng:

<kparal> is here anyone who knows how the pre-release warning in anaconda is configured? is it a releng config file?
<lsedlar> kparal, it's mostly automatic: when an RC compose is done, Pungi knows it's RC based on the compose label, and calls lorax with --isfinal option; which causes .buildstamp in the boot media to contain IsFinal=True and anaconda looks at that (I think)
<kparal> lsedlar: thanks. so how can we explain this? https://bugzilla.redhat.com/show_bug.cgi?id=1642286#c5
<lsedlar> kparal, I can't, I think even the workstation live should have it; maybe try asking someone from anaconda
<kparal> ok, thanks
<lsedlar> the .buildstamp in workstation initrd.img has IsFinal=False

On closer look this might not actually be a bug, rather confusion about the pre-release vs release/RC compose process for the DVD/netinst images.

According to my discussion with Fedora QA the difference between a pre-release and RC compose is:
- isFinal=True in .buildstamp
- the updates testing repo file in the fedora repos package has enabled=0

Also, as I've been told no such RC compose has yet been run.

Therefore the behavior the description talks about is apparently correct, as no RC compose has yet been done and any currently existing netinst/DVD image will be pre-release.

On the other hand, there seems to be an actual bug with the Live image, where Anaconda *does* behave as on a RC, even if it should not (as there is no RC yet).

Workstation Live is not related to this bug, because live images don't have the .buildstamp file. Anaconda uses workarounds to detect information about the release.

In case of the isFinal flag, Anaconda runs this piece of code:

RELEASE=$(rpm -q --qf '%{Release}' --whatprovides system-release)

if [ "${RELEASE:0:2}" = "0." ]; then
    export ANACONDA_ISFINAL="false"
else
    export ANACONDA_ISFINAL="true"
fi

These are results from Fedora-Workstation-Live-x86_64-29-20181022.n.1.iso:

$ rpm -q --whatprovides system-release
fedora-release-29-1.noarch

$ rpm -q --qf '%{Release}' --whatprovides system-release
1

$ echo "${RELEASE:0:2}"
1

The release is considered to be final.

Or in other words, once fedora-release has the release tag "-1" or greater, it's considered final, otherwise it's not, and that affects the pre-release warning on Live. I'm closing this as a NOTABUG, because this behavior has been with us for a long time (we just didn't notice before, I guess, or the timing with fedora-release was different). Anyway, shouldn't negatively affect the RC composes. We can talk about improving the anaconda+releng synchronization for the next releases.