Originally reported downstream in Gentoo at https://bugs.gentoo.org/948671.

We had a report of SQLite using the rdtsc x86 instruction on an i486 CPU (Soekris 4501 / AMD Elan) when SQLITE_ENABLE_STMT_SCANSTATUS is enabled.

While I appreciate that SQLITE_ENABLE_STMT_SCANSTATUS is rather unlikely to be used by anyone on such a CPU, we felt it was better to not vary the SQLite feature set based on ISA, so applied the following patch downstream:

From 0f2389e42ec65fd925e1bce608f382f3b35ee0f6 Mon Sep 17 00:00:00 2001
From: Sam James <sam@gentoo.org>
Date: Sat, 25 Jan 2025 12:10:24 +0000
Subject: [PATCH] hwtime.h: Don't use rdtsc on i486

If -DSQLITE_ENABLE_STMT_SCANSTATUS is passed, rdtsc is used for profiling
purposes, but that's not available on i486 CPUs (e.g. Soekris 4501 / AMD Elan
in the linked bug).

Just look for i586 instead of i386. For the i386/i486 case, it's fine if we
use the stub implementation. This feels cleaner than toggling functionality
in the ebuild, even if it's unlikely anyone on i486 is going to use that.

Thanks to amonakov who found the problem quickly and the relevant part of SQLite.

Bug: https://bugs.gentoo.org/948671
Signed-off-by: Sam James <sam@gentoo.org>
--- a/src/hwtime.h
+++ b/src/hwtime.h
@@ -24,7 +24,7 @@
 */
 #if !defined(__STRICT_ANSI__) && \
     (defined(__GNUC__) || defined(_MSC_VER)) && \
-    (defined(i386) || defined(__i386__) || defined(_M_IX86))
+    (defined(i586) || defined(__i586__) || defined(_M_IX86))

   #if defined(__GNUC__)

--
2.48.1

i586 should guarantee that rdtsc is available/usable.

As a style suggestion, it may be worth refactoring the conditional to pair the GCC check and then GCC macros, and then MSVC and MSVC macros to be a bit easier to read.