FYI, this has to wait for the F27 update to be pushed to stable first, which in turn has to wait for the -freeworld package to be ready. Please be patient.

FYI, the reason I am not pushing this to stable yet is that the -freeworld package is still being built, there are issues with the armv7hl build over there.

I really tried to get mesa fixed, but since they categorically refuse to apply the existing patches, Qt upstream gave up waiting and ended up adding this workaround. It makes me really sad, but with Nouveau upstream and downstream and FESCo conspiring to continue shipping a broken driver, there is no other option.

3.1.8 adds the following fixes:

• (bug) Netinstall would crash if the returned netinstall-groups data was empty.
• (regression vs. 3.0.x) UI glitch: netinstall would display the description as the name (ignoring the actual name) and show an empty description.
• (regression) GeoIP data had already been read, don't read it twice.
• Use more support code from KPMCore, instead of doing it ourselves.

3.1.8 adds the following fixes: * (bug) Netinstall would crash if the returned netinstall-groups data was empty. * (regression vs. 3.0.x) UI glitch: netinstall would display the description as the name (ignoring the actual name) and show an empty description. * (regression) GeoIP data had already been read, don't read it twice. * Use more support code from KPMCore, instead of doing it ourselves.

3.1.7 fixes the following regressions:

• Netinstall was no longer compatible with the packages module (regression in 3.1.4), which prevented any users of netinstall from installing packages from online repositories.
• The bootloader installation for EFI systems used the wrong name for the bootx64.efi (or bootia32.efi) files (regression in 3.1.1). The file was installed as grubx64.efi (or grubia32.efi). Some firmwares (in particular, the one in VirtualBox) are very picky about the name of that file and expect it to be bootx64.efi (or bootia32.efi), the installed system should now boot without tweaking on those systems again.

3.1.7 fixes the following regressions:

• Netinstall was no longer compatible with the packages module (regression in 3.1.4), which prevented any users of netinstall from installing packages from online repositories.
• The bootloader installation for EFI systems used the wrong name for the bootx64.efi (or bootia32.efi) files (regression in 3.1.1). The file was installed as grubx64.efi (or grubia32.efi). Some firmwares (in particular, the one in VirtualBox) are very picky about the name of that file and expect it to be bootx64.efi (or bootia32.efi), the installed system should now boot without tweaking on those systems again.

Shouldn't this obsolete libtxc_dxtn now that S3TC support is built in? Or are there other users of that library?

3.1.6 fixes a regression in 3.1.5: the finished.conf module configuration file was corrupt.

3.1.6 fixes a regression in 3.1.5: the finished.conf module configuration file was corrupt.

I've had it just not do anything. But this is with the old 5.8.0, mind you.

What I know is that Ctrl+f is not working right in plain text files even with QtWebEngine 5.8.0. So I don't think that that is a regression from this update.

The new build should address the issue reported in the above comments (https://bugreports.qt.io/browse/QTBUG-61521), please retest.

Looks like I have a fix that works now (built in Rawhide already): http://pkgs.fedoraproject.org/cgit/rpms/qt5-qtwebengine.git/plain/qtwebengine-opensource-src-5.9.0-qtbug-61521.patch?id=ace3e112e6b8e0338b84be9abe5e63d7052ecf54

I'll be updating the F26 update today.

This update has been unpushed.

Well, I was hoping to just edit the update with a fixed build, but now there is a mysterious further issue: https://bugzilla.redhat.com/show_bug.cgi?id=1464883#c2

See also https://bugzilla.redhat.com/show_bug.cgi?id=1464883 .