]> git.ipfire.org Git - thirdparty/gcc.git/commitdiff
testsuite: fix dwarf2/utf-1.C with DWARF4
authorMarek Polacek <polacek@redhat.com>
Wed, 5 Jul 2023 21:43:31 +0000 (17:43 -0400)
committerMarek Polacek <polacek@redhat.com>
Tue, 18 Jul 2023 17:01:30 +0000 (13:01 -0400)
Running
$ make check-c++ RUNTESTFLAGS='--target_board=unix\{-gdwarf-5,-gdwarf-4\} dwarf2.exp=utf-1.C'
shows
FAIL: g++.dg/debug/dwarf2/utf-1.C  -std=gnu++20  scan-assembler-times DW_AT_encoding \\(0x10\\) 3
because with -gdwarf-4 the output is:

  .byte   0x10    # DW_AT_encoding

but with -gdwarf-5 the output is the expected:

                # DW_AT_encoding (0x10)

The difference is caused by the DWARF5 optimize_implicit_const
optimization:
<https://gcc.gnu.org/pipermail/gcc-patches/2016-October/459762.html>

I suppose we could do what testsuite/rust/debug/chartype.rs does
and just run the test with -gdwarf-4.

gcc/testsuite/ChangeLog:

* g++.dg/debug/dwarf2/utf-1.C: Use -gdwarf-4.  Adjust expected
output.

gcc/testsuite/g++.dg/debug/dwarf2/utf-1.C

index 43b354f1bb5d1d4f6150f6003c661d080bdaa5ed..0ce4d8727d670bb75b7adb89537c2c72a2952c5a 100644 (file)
@@ -1,8 +1,13 @@
 // { dg-do compile { target c++20 } }
-// { dg-options { -gdwarf -dA } }
+// { dg-options { -gdwarf-4 -dA } }
 
 // Test that all three use DW_ATE_UTF.
-// { dg-final { scan-assembler-times {DW_AT_encoding \(0x10\)} 3 } }
+// This test uses -gdwarf-4 since in DWARF5 optimize_implicit_const
+// would optimize the output from:
+//   .byte   0x10    # DW_AT_encoding
+// into:
+//                   # DW_AT_encoding (0x10)
+// { dg-final { scan-assembler-times "0x10\[ \t]\[^\n\r]* DW_AT_encoding" 3 } }
 
 char8_t c8;
 char16_t c16;