Fix the false -Warray-bounds error by setting the type after the value

The pragma around set_parents() is no longer needed: the base branch
(#5585) creates a value before setting its type. Before, the type of a new
string was stored first, so GCC had to assume that operator new could
change it again, kept the object branch of the inlined set_parents() alive,
and checked it against the string's allocation. Rewriting set_parents()
itself (if/else instead of switch) does not help.

Verified with Docker GCC 12.5, 13, 14, 15.3 and 16.2: the regression test
builds and passes without the pragma, and fails to build on develop.

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

include/nlohmann/json.hpp

@@ -763,15 +763,6 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
         static_cast<void>(check_parents);
     }
 
-    // GCC 13 to at least 15 report a false -Warray-bounds error when set_parents()
-    // is inlined at -O3 right after a non-container value was created: the analysis
-    // does not use m_type to rule out the object/array branches and checks
-    // the std::map access against the allocation of, e.g., a string.
-    // See https://github.com/nlohmann/json/issues/5742 and #4819.
-#if defined(__GNUC__) && !defined(__clang__)
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Warray-bounds"
-#endif
     void set_parents()
     {
 #if JSON_DIAGNOSTICS
@@ -808,9 +799,6 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
         }
 #endif
     }
-#if defined(__GNUC__) && !defined(__clang__)
-#pragma GCC diagnostic pop
-#endif
 
     iterator set_parents(iterator it, std::ptrdiff_t count_set_parents)
     {

single_include/nlohmann/json.hpp

@@ -27461,15 +27461,6 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
         static_cast<void>(check_parents);
     }
 
-    // GCC 13 to at least 15 report a false -Warray-bounds error when set_parents()
-    // is inlined at -O3 right after a non-container value was created: the analysis
-    // does not use m_type to rule out the object/array branches and checks
-    // the std::map access against the allocation of, e.g., a string.
-    // See https://github.com/nlohmann/json/issues/5742 and #4819.
-#if defined(__GNUC__) && !defined(__clang__)
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Warray-bounds"
-#endif
     void set_parents()
     {
 #if JSON_DIAGNOSTICS
@@ -27506,9 +27497,6 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
         }
 #endif
     }
-#if defined(__GNUC__) && !defined(__clang__)
-#pragma GCC diagnostic pop
-#endif
 
     iterator set_parents(iterator it, std::ptrdiff_t count_set_parents)
     {

tests/src/unit-diagnostics-optimized.cpp

@@ -7,10 +7,14 @@
 // SPDX-License-Identifier: MIT
 
 // Regression test for https://github.com/nlohmann/json/issues/5742: with
-// JSON_DIAGNOSTICS, GCC (13 to at least 15) reports a false -Warray-bounds
-// error in the inlined set_parents() at -O3. The warning depends on GCC's
-// inlining decisions, so the sections cover patterns that trigger it on
-// different GCC versions (#4819: GCC 13 and 14; #5742: GCC 14 and 15).
+// JSON_DIAGNOSTICS, GCC (12 to at least 16) reported a false -Warray-bounds
+// error in the inlined set_parents() at -O3. The type of a new string was set
+// before the string was allocated, so GCC had to assume that operator new
+// could change it again and checked the object branch of set_parents()
+// against the string's allocation. Setting the type after creating the value
+// avoids this. The warning depends on GCC's inlining decisions, so the
+// sections cover two patterns that trigger it on different GCC versions
+// (#4819 and #5742).
 // On GCC, this file is compiled with -O3 -Werror=array-bounds (see
 // tests/CMakeLists.txt), so the test fails to build if the warning returns.