Summary: | WebSocketChannel.h build failure with -Werror=overloaded-virtual on GCC 13 | ||
---|---|---|---|
Product: | WebKit | Reporter: | Fujii Hironori <Hironori.Fujii> |
Component: | WebKit2 | Assignee: | Nobody <webkit-unassigned> |
Status: | RESOLVED FIXED | ||
Severity: | Normal | CC: | cadubentzen, csaavedra, darin, kkinnunen, webkit-bug-importer |
Priority: | P2 | Keywords: | InRadar |
Version: | WebKit Nightly Build | ||
Hardware: | Unspecified | ||
OS: | Unspecified |
Description
Fujii Hironori
2024-04-22 17:40:43 PDT
Is it possible that this is caused by unified builds? I can't reproduce this by building WKGTK from directly from git, but when I build a couple of my own MRs, I hit this locally. IIUC we have virtual bool MessageSender::sendMessage(UniqueRef<Encoder>&&, OptionSet<SendOption>); and then template<typename T> void WebSocketChannel::sendMessage(T&&, size_t byteLength); pulled into the same compilation unit. WebSocketChannel inherits from MessageSender: class WebSocketChannel : public IPC::MessageSender, public IPC::MessageReceiver, public WebCore::ThreadableWebSocketChannel, public RefCounted<WebSocketChannel> { Is this intentional? It may be because of -Woverloaded-virtual being included in -Wall on GCC 13: https://gcc.gnu.org/git/?p=gcc.git;a=commit;h=113844d68e94f4e9c0e946db351ba7d3d4a1335a Pull request: https://github.com/WebKit/WebKit/pull/27625 Not GTK-specific, maybe GCC-specific? The title could be set to something like `Build failure with -Werror=overloaded-virtual in GCC`. Would be nice to notice things like this in EWS. Maybe we can update the GCC used in the GTK EWS? Something weird is happening on GTK EWS. I'm using the latest flatpak SDK. It actually contains GCC 13.2.0. > fujii@fujihiro-ubuntu $ ./Tools/Scripts/webkit-flatpak --version > 277469@main > fujii@fujihiro-ubuntu $ ./Tools/Scripts/webkit-flatpak -c /usr/bin/c++ --version > c++ (GCC) 13.2.0 GTK EWS workers are also claiming they are using "SDK version: 277469@main". At least for me, this only happens after applying a couple of changes in Source/WebKit locally and building. When building a pristine revision from git, it builds fine. So I don't think it's a EWS issue. Could it be caused by varying grouping of sources in each unified build file? Different unified build grouping only happens when we add or remove source files, I believe. Committed 277904@main (619e39d2eab1): <https://commits.webkit.org/277904@main> Reviewed commits have been landed. Closing PR #27625 and removing active labels. |