How does the given patch fix this?

ACK ef11f39

Tested that I can compile with --enable-debug. Also no idea why this works.

Could use some documentation as to why ef11f39 fixes the issue, but it does.

How does the given patch fix this?

Not sure ((

It seems related to instantiating of the FatalError() template function, and all used symbols should be in the same translation unit. But don't know how this depends on --enable-debug.

FWIW, I've managed to reproduce the initial issue with the only flag:

$ ./configure CXXFLAGS=-O0

UPDATE: using GCC 7.5.0

The alternative approach is not using function aliasing at all:

diff --git a/src/ui_interface.cpp b/src/ui_interface.cpp
index 15795bd67..c6017e57a 100644
--- a/src/ui_interface.cpp
+++ b/src/ui_interface.cpp
@@ -59,6 +59,11 @@ bool InitError(const bilingual_str& str)
     return false;
 }
 
+bool AbortError(const bilingual_str& str)
+{
+    return InitError(str);
+}
+
 void InitWarning(const bilingual_str& str)
 {
     uiInterface.ThreadSafeMessageBox(str, "", CClientUIInterface::MSG_WARNING);
diff --git a/src/ui_interface.h b/src/ui_interface.h
index 356d30eaf..bfa771a1b 100644
--- a/src/ui_interface.h
+++ b/src/ui_interface.h
@@ -122,7 +122,7 @@ void InitWarning(const bilingual_str& str);
 
 /** Show error message **/
 bool InitError(const bilingual_str& str);
-constexpr auto AbortError = InitError;
+bool AbortError(const bilingual_str& str);
 
 extern CClientUIInterface uiInterface;
 

Updated ef11f39 -> b83cc0f (pr19309.01 -> pr19309.02, diff):

• addressed @MarcoFalke's comment:

A smaller patch would be to write the verbose version of the alias?

Is there an upstream GCC bug for this? If we decide to go with this workaround, perhaps leave a comment with some context and we can revert in the future?

@dongcarl I don't think anyone has reduced the test case yet

Re-ACK b83cc0f

This is not a gcc bug, but a Bitcoin Core bug. Proper fix is here: #19331

@MarcoFalke

This is not a gcc bug, but a Bitcoin Core bug. Proper fix is here: #19331

Thank you Marco!