Failed to verify signature archive-contents.sig:
No public key for 645357D2883A0966 created at 2026-05-12T17:10:22-0400 using EDDSA
Command output:
gpg: keyblock resource '/c/Users/Matty/Documents/Emacs/c:/Users/Matty/.emacs.d/elpa/gnupg/pubring.kbx': No such file or directory
gpg: Signature made Tue May 12 17:10:22 2026 EST
gpg: using EDDSA key 0327BE68D64D9A1A66859F15645357D2883A0966
gpg: Can't check signature: No public key
The obvious issue is that Emacs is looking at a nonsense path:
/c/Users/Matty/Documents/Emacs/c:/Users/Matty/.emacs.d/elpa/gnupg/pubring.kbx
which is clearly the concatenation of two different paths.
But note that the two paths start differently:
/c/c:/The second path is in typical Windows-style, but the first path is something you get when Windows is pretending to be Linux
default-directoryChecking the value of my default-directory variable
Its value is "~/Documents/Emacs/"
Local in buffer *scratch*; global value is "C:/Users/Matty/"
It's a bit weird that the default directory has a local and global value, no?
In Emacs, default-directory is
buffer-local, meaning
*scratch* buffer) will have its local default
directory set to the folder where the program was launched.setq (Set
Quote)Sets the value of a variable in the current buffer if that variable is buffer-local.
setq-defaultChanges the default value for all new buffers, but ignores the current buffer
All the stuff I tried before was going off in the wrong direction.
Instead, using a relative path for my package-gnupghome-dir global
variable was enough:
(setq package-gnupghome-dir "~/.emacs.d/elpa/gnupg")Somehow absolute paths cause double-rooting in the GPG on Windows 11
Git for Windows uses a specialized MSYS2 environment that prefers the SQLite (.db) backend, while Emacs expects the standard GNU/Linux file-based structure (.kbx).
By installing a "Vanilla" GnuPG binary into your personal bin/ folder, you create a deterministic environment that you control, bypassing the "Shadow Path" issues and the MSYS2 translation layer.
The "Staff Engineer" Autopsy: Why it works now
While the specific fix feels like "magic," we actually addressed four distinct architectural layers:
The Cryptographic Anchor: By setting GNUPGHOME and the gpg.exe path before initializing packages, you ensured the "handshake" didn't fail at the hardware layer.
The Pipe Fix: process-connection-type nil solved the Windows-specific "Invalid Argument" error by switching from a PTY to a simpler stream that Windows 11 actually understands.
The Dependency Reset: Deleting the elpa/ folders removed "poisoned" files that were byte-compiled incorrectly during the earlier failures.
The Built-in Override: package-install-upgrade-built-in allowed Magit to "win" its conflict with the older transient version that ships with Emacs.