Yesterday I upgraded from Fedora 42 to 44 and I've just now realized that QuiteRSS has broken. It doesn't render articles any more. The right side lower pane is just empty. There are no errors logged in console. It's basically a dupe of bug 2457248 but I'm filing it nonetheless because I'm 99.99% confident it's a bug in Qt5 rather than the application itself that has seen no releases or commits for years now.
Created attachment 2136563 [details] Empty preview pane Just to make things clear. The built-in web browser still works. It's just the preview pane that's empty.
I've replaced qt5-qtwebkit-5.212.0-0.98alpha4.fc44.x86_64 with qt5-qtwebkit-0:5.212.0-0.95alpha4.fc42.x86_64 and now the app is working fine. Something in between 0.95 and 0.98 has broken rendering in QuiteRSS.
If someone else is still using QuiteRSS here's how you can fix it: 1. Download qt5-qtwebkit-5.212.0-0.95alpha4.fc43.x86_64.rpm from any Fedora mirror. 2. Run rpm -K qt5-qtwebkit-5.212.0-0.95alpha4.fc43.x86_64.rpm it should return: qt5-qtwebkit-5.212.0-0.95alpha4.fc43.x86_64.rpm: digests signatures OK 3. Now run sudo rpm -Uvh --force qt5-qtwebkit-5.212.0-0.95alpha4.fc43.x86_64.rpm 4. Restart QuiteRSS. You're all set.
Also, I have no note that the Fedora 44 binaries have become substantially fatter: Fedora 43: -rwxr-xr-x 1 root root 45298584 Nov 4 04:00 /usr/lib64/libQt5WebKit.so.5.212.0 Fedora 44: -rwxr-xr-x 1 root root 47444128 Feb 7 04:00 /usr/lib64/libQt5WebKit.so.5.212.0 4.7% bigger. Is this due to a new GCC? Or new compilation flags? But why?
There is one final, very important detail. The articles are indeed rendered, contrary to what I wrote before. However, it appears that the dimensions of the preview pane are 150,000 x 150,000 pixels, and the text and image of the article are rendered in the bottom left corner. I've tried adjusting the zoom level, but while the preview pane shrinks, the text simply becomes unreadable and the dimensions are still far too big. I'm not a programmer, but it seems that some integer logic has broken down.
This very much looks like a regression caused by the switch to the GCC 16 prerelease (gcc-16.0.1-0.10.fc44) used for the F44 mass rebuild. The previous working version (0.95alpha4.fc43) was built with stable GCC 15.2.1, and no equivalent rebuild exists for F43.
Is it possible to recompile this package using just released GCC 16.1?
It seems with new qt5-qtwebkit you need to double click on your rss news in the list. It should work. Please try!
(In reply to Than Ngo from comment #8) Depending on your settings, double-clicking either opens a **full-featured** built-in web browser or an external web browser, as is the case for me. Viewing or checking multiple articles this way is extremely cumbersome, tiresome and counterintuitive. In any case, it looks like you've just confirmed this regression, and I'm wondering if I need to involve the GCC team because I still think it's a regression caused by the transition from GCC 15 to GCC 16.
If you suspect miscompilation of qt5-qtwebkit (or some dependency thereof), can you please help to narrow it down? If F43 and F44 are using roughly same version, can you try to replace some shared libraries or binaries from the known working F43 version to find out which exact library or binary it is? Is it built with LTO? If yes, can you reproduce it with %global _lto_cflags %{nil} ? If yes (i.e. it is reproduceable without LTO), ideally try to build the library or binary side by side in F44 and F43 chroots and then try to bisect among the object files which one is problematic? Usually start with half *.o files from the known good set and another from the possibly bad set, etc. If LTO is needed, then try to keep removing -flto from sources as long as it does reproduce (can be done also by doing a LTO and non-LTO builds side by side and bisect among the object files from those two builds.
Replacing this library **alone** (with the version from Fedora 43) makes QuiteRSS work flawlessly, so it's only the qt5-qtwebkit component that's being miscompiled and nothing else. On the GCC side they have confirmed GCC 16 does miscompile applications here and there.
And according to this: https://bugs.gentoo.org/show_bug.cgi?id=969496 >The broken rendering seems to show issues with the width and heights of elements. Liferea shows a big black rectangle on the right of the "Unread" pane. webkit-gtk's 6 SLOT is probably affected too, but I haven't tried. which matches this bug report almost verbatim. webkit-gtk is equally affected.
Lastly: >I've now tried 2.50.6-r410 against 16.0.1_p20260412. The problem seems to have gone away. :) Can you please recompile qt5-qtwebkit using GCC 16.1 and push it to testing? Wait, you've not pushed GCC 16.1 yet. Please do.
I can't until karma is given: https://bodhi.fedoraproject.org/updates/FEDORA-2026-1f4ff51dc3
It is now in f44 and f45 buildroots (for 2 days already).
(In reply to Jakub Jelinek from comment #15) Thanks, I wasn't paying attention to this package. I've just installed qt5-qtwebkit-5.212.0-0.98alpha4.fc44.x86_64.rpm and the issue sadly persists :-( I'm closing this bug report because it's related to upstream and I don't think Fedora can realistically do anything to fix it unless somehow you manage to install GCC 15.x in a Fedora 44 chroot to (re)build it. If people stumble on this bug [report] they will need to install qt5-qtwebkit-5.212.0-0.95alpha4.fc43.x86_64 and not upgrade to the FC44 version of it.
I'm just curious. Is Fedora 44's set of compiler and linker flags the same as Fedora's 43? I've no idea how to check it.