Fix test_bitcoin-qt when DISPLAY is unset#10113
Fix test_bitcoin-qt when DISPLAY is unset#10113ryanofsky wants to merge 1 commit intobitcoin:masterfrom
Conversation
Issue reported by Matt Corallo <git@bluematt.me> in bitcoin#10110 Fixes bitcoin#10110
| #endif | ||
|
|
||
| static bool UsesXcb() { | ||
| for (QObject* plugin: QPluginLoader::staticInstances()) { |
There was a problem hiding this comment.
How does this react if xcb is available but not default? For ex, if egl or wayland would otherwise be used?
There was a problem hiding this comment.
How does this react if xcb is available but not default? For ex, if egl or wayland would otherwise be used?
This is a good point, I suspect the check is broken in this case.
|
I think the |
ryanofsky
left a comment
There was a problem hiding this comment.
I think the make check tests should never try to use the DISPLAY (or the computer's GUI otherwise, e.g. through egl/wayland/win32/quartz/...).
There's a place for tests that require rendering (usually combined with a "monkey" to mimic user interaction) but these should IMO be separate when we have them.
The test doesn't actually require any rendering, it's just failing to initialize QApplication (which must be calling some libxcb functions) and crashing on startup.
I'm closing this in favor of an alternative fix in #10117 though, which modifies "make check" to run the test under xvfb if it's available.
| #endif | ||
|
|
||
| static bool UsesXcb() { | ||
| for (QObject* plugin: QPluginLoader::staticInstances()) { |
There was a problem hiding this comment.
How does this react if xcb is available but not default? For ex, if egl or wayland would otherwise be used?
This is a good point, I suspect the check is broken in this case.
Issue reported by @TheBlueMatt in #10110