Remove unnecessary comparison.

Thank you, Victor, for verifying that clang is able to determine at compile time
that |value| can never be larger than numeric_limits::max() and therefore is
able to optimize away the if branch.  See cr/227019741.

gfe-relnote: n/a Should not change binary.
PiperOrigin-RevId: 227033117
Change-Id: Icf319bf50837c3843986a102d1232879d655dc45
diff --git a/http2/hpack/decoder/hpack_string_decoder.h b/http2/hpack/decoder/hpack_string_decoder.h
index 88b1add..5e0d220 100644
--- a/http2/hpack/decoder/hpack_string_decoder.h
+++ b/http2/hpack/decoder/hpack_string_decoder.h
@@ -173,10 +173,8 @@
     const uint64_t value = length_decoder_.value();
     // |remaining_| is size_t.  Check for truncation on 32-bit platforms.
     // numeric_limits::max() is constexpr.  On platforms where size_t is at
-    // least 64 bit wide, the compiler optimizes away this branch.
-    if (std::numeric_limits<uint64_t>::max() >
-            std::numeric_limits<size_t>::max() &&
-        value > std::numeric_limits<size_t>::max()) {
+    // least 64 bit wide, the compiler should optimize away this branch.
+    if (value > std::numeric_limits<size_t>::max()) {
       *status = DecodeStatus::kDecodeError;
       return;
     }