UBSAN has found me a reliance on uninitialised data / undefined
behaviour that affects these keys specifically. Fixed here:
https://git.tartarus.org/?p=simon/putty.git;a=commitdiff;h=6a88b294276b9c24584efa18b9a37f437fa46712
This may have some bearing on the matter.
If someone (who can reproduce this behaviour with a freshly locally
compiled version of PuTTY/pterm) can try applying this patch (attached
for convenience) and see if it helps, that might be helpful.
(If this is it, it's been latent since 0.78; I guess we just started
getting unlucky with the toolchain recently or something.)
commit 6a88b294276b9c24584efa18b9a37f437fa46712
Author: Jacob Nevins <jac...@chiark.greenend.org.uk>
Date: Wed Dec 4 17:21:30 2024 +0000
Unix PuTTY/pterm: fix UB with small keypad.
We were relying on uninitialised data. Found by UBSAN.
(Introduced in commit c88b6d1853, I think.)
diff --git a/unix/window.c b/unix/window.c
index db0eff31..4a8968d6 100644
--- a/unix/window.c
+++ b/unix/window.c
@@ -1927,6 +1927,7 @@ gint key_event(GtkWidget *widget, GdkEventKey *event, gpointer data)
if (event->state & GDK_CONTROL_MASK)
break;
+ consumed_meta_key = false;
end = 1 + format_small_keypad_key(
output+1, inst->term, sk_key, event->state & GDK_SHIFT_MASK,
event->state & GDK_CONTROL_MASK,
