[libc++] Fix num_get base parsing

[philnik777]

This fixes two bugs reported in #121795 and adds regression tests. Specifically, these bugs are in the base detection mechanism. The first bug is that the out parameter isn't set when the stream only contains zero and after that is the end of the stream. The second one is that we don't consider 0 to be a number, and instead we only parse it as the start of an octal literal.

[github-actions]

⚠️ C/C++ code formatter, clang-format found issues in your code. ⚠️

You can test this locally with the following command:

git-clang-format --diff origin/main HEAD --extensions h,cpp -- libcxx/include/__locale_dir/num.h libcxx/test/std/localization/locale.categories/category.numeric/locale.num.get/facet.num.get.members/get_long.pass.cpp --diff_from_common_commit

⚠️
The reproduction instructions above might return results for more than one PR
in a stack if you are using a stacked PR workflow. You can limit the results by
changing origin/main to the base branch/commit you want to compare against.
⚠️

View the diff from clang-format here.

diff --git a/libcxx/test/std/localization/locale.categories/category.numeric/locale.num.get/facet.num.get.members/get_long.pass.cpp b/libcxx/test/std/localization/locale.categories/category.numeric/locale.num.get/facet.num.get.members/get_long.pass.cpp
index 45b034833..db9c4585d 100644
--- a/libcxx/test/std/localization/locale.categories/category.numeric/locale.num.get/facet.num.get.members/get_long.pass.cpp
+++ b/libcxx/test/std/localization/locale.categories/category.numeric/locale.num.get/facet.num.get.members/get_long.pass.cpp
@@ -670,101 +670,101 @@ int main(int, char**)
       assert(v == std::numeric_limits<long>::min());
     }
 
-  { // Check that auto-detection of the base works properly
-    ios.flags(ios.flags() & ~std::ios::basefield);
-    { // zeroes
-      {
-        v                          = -1;
-        const char str[]           = "0";
-        std::ios_base::iostate err = ios.goodbit;
+    { // Check that auto-detection of the base works properly
+      ios.flags(ios.flags() & ~std::ios::basefield);
+      { // zeroes
+        {
+          v                          = -1;
+          const char str[]           = "0";
+          std::ios_base::iostate err = ios.goodbit;
 
-        cpp17_input_iterator<const char*> iter =
-            f.get(cpp17_input_iterator<const char*>(str), cpp17_input_iterator<const char*>(str + 1), ios, err, v);
-        assert(base(iter) == str + 1);
-        assert(err == ios.eofbit);
-        assert(v == 0);
-      }
-      {
-        v                          = -1;
-        const char str[]           = "00";
-        std::ios_base::iostate err = ios.goodbit;
+          cpp17_input_iterator<const char*> iter =
+              f.get(cpp17_input_iterator<const char*>(str), cpp17_input_iterator<const char*>(str + 1), ios, err, v);
+          assert(base(iter) == str + 1);
+          assert(err == ios.eofbit);
+          assert(v == 0);
+        }
+        {
+          v                          = -1;
+          const char str[]           = "00";
+          std::ios_base::iostate err = ios.goodbit;
 
-        cpp17_input_iterator<const char*> iter =
-            f.get(cpp17_input_iterator<const char*>(str), cpp17_input_iterator<const char*>(str + 2), ios, err, v);
-        assert(base(iter) == str + 2);
-        assert(err == ios.eofbit);
-        assert(v == 0);
-      }
-      {
-        v                          = -1;
-        const char str[]           = "0x0";
-        std::ios_base::iostate err = ios.goodbit;
+          cpp17_input_iterator<const char*> iter =
+              f.get(cpp17_input_iterator<const char*>(str), cpp17_input_iterator<const char*>(str + 2), ios, err, v);
+          assert(base(iter) == str + 2);
+          assert(err == ios.eofbit);
+          assert(v == 0);
+        }
+        {
+          v                          = -1;
+          const char str[]           = "0x0";
+          std::ios_base::iostate err = ios.goodbit;
 
-        cpp17_input_iterator<const char*> iter =
-            f.get(cpp17_input_iterator<const char*>(str), cpp17_input_iterator<const char*>(str + 3), ios, err, v);
-        assert(base(iter) == str + 3);
-        assert(err == ios.eofbit);
-        assert(v == 0);
-      }
-      {
-        v                          = -1;
-        const char str[]           = "0X0";
-        std::ios_base::iostate err = ios.goodbit;
+          cpp17_input_iterator<const char*> iter =
+              f.get(cpp17_input_iterator<const char*>(str), cpp17_input_iterator<const char*>(str + 3), ios, err, v);
+          assert(base(iter) == str + 3);
+          assert(err == ios.eofbit);
+          assert(v == 0);
+        }
+        {
+          v                          = -1;
+          const char str[]           = "0X0";
+          std::ios_base::iostate err = ios.goodbit;
 
-        cpp17_input_iterator<const char*> iter =
-            f.get(cpp17_input_iterator<const char*>(str), cpp17_input_iterator<const char*>(str + 3), ios, err, v);
-        assert(base(iter) == str + 3);
-        assert(err == ios.eofbit);
-        assert(v == 0);
+          cpp17_input_iterator<const char*> iter =
+              f.get(cpp17_input_iterator<const char*>(str), cpp17_input_iterator<const char*>(str + 3), ios, err, v);
+          assert(base(iter) == str + 3);
+          assert(err == ios.eofbit);
+          assert(v == 0);
+        }
       }
-    }
-    { // first character after base is out of range
-      {
-        v                          = -1;
-        const char str[]           = "08";
-        std::ios_base::iostate err = ios.goodbit;
+      { // first character after base is out of range
+        {
+          v                          = -1;
+          const char str[]           = "08";
+          std::ios_base::iostate err = ios.goodbit;
 
-        cpp17_input_iterator<const char*> iter =
-            f.get(cpp17_input_iterator<const char*>(str), cpp17_input_iterator<const char*>(str + 2), ios, err, v);
-        assert(base(iter) == str + 1);
-        assert(err == ios.goodbit);
-        assert(v == 0);
-      }
-      {
-        v                          = -1;
-        const char str[]           = "1a";
-        std::ios_base::iostate err = ios.goodbit;
+          cpp17_input_iterator<const char*> iter =
+              f.get(cpp17_input_iterator<const char*>(str), cpp17_input_iterator<const char*>(str + 2), ios, err, v);
+          assert(base(iter) == str + 1);
+          assert(err == ios.goodbit);
+          assert(v == 0);
+        }
+        {
+          v                          = -1;
+          const char str[]           = "1a";
+          std::ios_base::iostate err = ios.goodbit;
 
-        cpp17_input_iterator<const char*> iter =
-            f.get(cpp17_input_iterator<const char*>(str), cpp17_input_iterator<const char*>(str + 2), ios, err, v);
-        assert(base(iter) == str + 1);
-        assert(err == ios.goodbit);
-        assert(v == 1);
-      }
-      {
-        v                          = -1;
-        const char str[]           = "0xg";
-        std::ios_base::iostate err = ios.goodbit;
+          cpp17_input_iterator<const char*> iter =
+              f.get(cpp17_input_iterator<const char*>(str), cpp17_input_iterator<const char*>(str + 2), ios, err, v);
+          assert(base(iter) == str + 1);
+          assert(err == ios.goodbit);
+          assert(v == 1);
+        }
+        {
+          v                          = -1;
+          const char str[]           = "0xg";
+          std::ios_base::iostate err = ios.goodbit;
 
-        cpp17_input_iterator<const char*> iter =
-            f.get(cpp17_input_iterator<const char*>(str), cpp17_input_iterator<const char*>(str + 3), ios, err, v);
-        assert(base(iter) == str + 2);
-        assert(err == ios.failbit);
-        assert(v == 0);
-      }
-      {
-        v                          = -1;
-        const char str[]           = "0Xg";
-        std::ios_base::iostate err = ios.goodbit;
+          cpp17_input_iterator<const char*> iter =
+              f.get(cpp17_input_iterator<const char*>(str), cpp17_input_iterator<const char*>(str + 3), ios, err, v);
+          assert(base(iter) == str + 2);
+          assert(err == ios.failbit);
+          assert(v == 0);
+        }
+        {
+          v                          = -1;
+          const char str[]           = "0Xg";
+          std::ios_base::iostate err = ios.goodbit;
 
-        cpp17_input_iterator<const char*> iter =
-            f.get(cpp17_input_iterator<const char*>(str), cpp17_input_iterator<const char*>(str + 3), ios, err, v);
-        assert(base(iter) == str + 2);
-        assert(err == ios.failbit);
-        assert(v == 0);
+          cpp17_input_iterator<const char*> iter =
+              f.get(cpp17_input_iterator<const char*>(str), cpp17_input_iterator<const char*>(str + 3), ios, err, v);
+          assert(base(iter) == str + 2);
+          assert(err == ios.failbit);
+          assert(v == 0);
+        }
       }
     }
-  }
 
   return 0;
 }

[llvmbot]

@llvm/pr-subscribers-libcxx

Author: Nikolas Klauser (philnik777)

Changes

---

Full diff: https://github.com/llvm/llvm-project/pull/170460.diff

2 Files Affected:

• (modified) libcxx/include/__locale_dir/num.h (+2)
• (modified) libcxx/test/std/localization/locale.categories/category.numeric/locale.num.get/facet.num.get.members/get_long.pass.cpp (+96)

diff --git a/libcxx/include/__locale_dir/num.h b/libcxx/include/__locale_dir/num.h
index 98b8eb0c600f5..b7ea02e7cb7f7 100644
--- a/libcxx/include/__locale_dir/num.h
+++ b/libcxx/include/__locale_dir/num.h
@@ -436,6 +436,7 @@ class num_get : public locale::facet, private __num_get<_CharT> {
         ++__first;
         if (__first == __last) {
           __err |= ios_base::eofbit;
+          __v = 0;
           return __first;
         }
         // __c2 == 'x' || __c2 == 'X'
@@ -444,6 +445,7 @@ class num_get : public locale::facet, private __num_get<_CharT> {
           ++__first;
         } else {
           __base = 8;
+          __parsed_num = true; // We only swallowed '0', so we've started to parse a number
         }
       } else {
         __base = 10;
diff --git a/libcxx/test/std/localization/locale.categories/category.numeric/locale.num.get/facet.num.get.members/get_long.pass.cpp b/libcxx/test/std/localization/locale.categories/category.numeric/locale.num.get/facet.num.get.members/get_long.pass.cpp
index a110aae2db11b..45b034833fecc 100644
--- a/libcxx/test/std/localization/locale.categories/category.numeric/locale.num.get/facet.num.get.members/get_long.pass.cpp
+++ b/libcxx/test/std/localization/locale.categories/category.numeric/locale.num.get/facet.num.get.members/get_long.pass.cpp
@@ -670,5 +670,101 @@ int main(int, char**)
       assert(v == std::numeric_limits<long>::min());
     }
 
+  { // Check that auto-detection of the base works properly
+    ios.flags(ios.flags() & ~std::ios::basefield);
+    { // zeroes
+      {
+        v                          = -1;
+        const char str[]           = "0";
+        std::ios_base::iostate err = ios.goodbit;
+
+        cpp17_input_iterator<const char*> iter =
+            f.get(cpp17_input_iterator<const char*>(str), cpp17_input_iterator<const char*>(str + 1), ios, err, v);
+        assert(base(iter) == str + 1);
+        assert(err == ios.eofbit);
+        assert(v == 0);
+      }
+      {
+        v                          = -1;
+        const char str[]           = "00";
+        std::ios_base::iostate err = ios.goodbit;
+
+        cpp17_input_iterator<const char*> iter =
+            f.get(cpp17_input_iterator<const char*>(str), cpp17_input_iterator<const char*>(str + 2), ios, err, v);
+        assert(base(iter) == str + 2);
+        assert(err == ios.eofbit);
+        assert(v == 0);
+      }
+      {
+        v                          = -1;
+        const char str[]           = "0x0";
+        std::ios_base::iostate err = ios.goodbit;
+
+        cpp17_input_iterator<const char*> iter =
+            f.get(cpp17_input_iterator<const char*>(str), cpp17_input_iterator<const char*>(str + 3), ios, err, v);
+        assert(base(iter) == str + 3);
+        assert(err == ios.eofbit);
+        assert(v == 0);
+      }
+      {
+        v                          = -1;
+        const char str[]           = "0X0";
+        std::ios_base::iostate err = ios.goodbit;
+
+        cpp17_input_iterator<const char*> iter =
+            f.get(cpp17_input_iterator<const char*>(str), cpp17_input_iterator<const char*>(str + 3), ios, err, v);
+        assert(base(iter) == str + 3);
+        assert(err == ios.eofbit);
+        assert(v == 0);
+      }
+    }
+    { // first character after base is out of range
+      {
+        v                          = -1;
+        const char str[]           = "08";
+        std::ios_base::iostate err = ios.goodbit;
+
+        cpp17_input_iterator<const char*> iter =
+            f.get(cpp17_input_iterator<const char*>(str), cpp17_input_iterator<const char*>(str + 2), ios, err, v);
+        assert(base(iter) == str + 1);
+        assert(err == ios.goodbit);
+        assert(v == 0);
+      }
+      {
+        v                          = -1;
+        const char str[]           = "1a";
+        std::ios_base::iostate err = ios.goodbit;
+
+        cpp17_input_iterator<const char*> iter =
+            f.get(cpp17_input_iterator<const char*>(str), cpp17_input_iterator<const char*>(str + 2), ios, err, v);
+        assert(base(iter) == str + 1);
+        assert(err == ios.goodbit);
+        assert(v == 1);
+      }
+      {
+        v                          = -1;
+        const char str[]           = "0xg";
+        std::ios_base::iostate err = ios.goodbit;
+
+        cpp17_input_iterator<const char*> iter =
+            f.get(cpp17_input_iterator<const char*>(str), cpp17_input_iterator<const char*>(str + 3), ios, err, v);
+        assert(base(iter) == str + 2);
+        assert(err == ios.failbit);
+        assert(v == 0);
+      }
+      {
+        v                          = -1;
+        const char str[]           = "0Xg";
+        std::ios_base::iostate err = ios.goodbit;
+
+        cpp17_input_iterator<const char*> iter =
+            f.get(cpp17_input_iterator<const char*>(str), cpp17_input_iterator<const char*>(str + 3), ios, err, v);
+        assert(base(iter) == str + 2);
+        assert(err == ios.failbit);
+        assert(v == 0);
+      }
+    }
+  }
+
   return 0;
 }

[alexfh]

This fixes the problems we've seen so far. Thank you!

[alexfh]

LG. Thanks! This resolves the problems for us.