format-test fails 3 tests under GCC with -std=c++11 rather than -std=gnu++11

[tonyelewis]

Building and running format-test on Ubuntu 20.04 with:

rm -rf build
mkdir build
cmake -GNinja -Bbuild -H. -DCMAKE_CXX_EXTENSIONS=OFF
ninja -C build format-test
build/bin/format-test

…I get three failures:

[...]

[ RUN      ] FormatterTest.FormatBin
[       OK ] FormatterTest.FormatBin (0 ms)
[ RUN      ] FormatterTest.FormatDec
../test/format-test.cc:1124: Failure
Value of: format("{0}", static_cast<__int128_t>((9223372036854775807L)) + 1)
  Actual: "0"
Expected: "9223372036854775808"
../test/format-test.cc:1126: Failure
Value of: format("{0}", static_cast<__int128_t>((-9223372036854775807L-1)) - 1)
  Actual: "4294967295"
Expected: "-9223372036854775809"
../test/format-test.cc:1128: Failure
Value of: format("{0}", static_cast<__int128_t>((18446744073709551615UL)) + 1)
  Actual: "0"
Expected: "18446744073709551616"
../test/format-test.cc:1130: Failure
Value of: format("{0}", int128_max)
  Actual: "4294967295"
Expected: "170141183460469231731687303715884105727"
../test/format-test.cc:1132: Failure
Value of: format("{0}", int128_min)
  Actual: "0"
Expected: "-170141183460469231731687303715884105728"
../test/format-test.cc:1134: Failure
Value of: format("{0}", uint128_max)
  Actual: "4294967295"
Expected: "340282366920938463463374607431768211455"
[  FAILED  ] FormatterTest.FormatDec (0 ms)
[ RUN      ] FormatterTest.FormatHex
../test/format-test.cc:1165: Failure
Value of: format("{0:x}", static_cast<__int128_t>((9223372036854775807L)) + 1)
  Actual: "0"
Expected: "8000000000000000"
../test/format-test.cc:1167: Failure
Value of: format("{0:x}", static_cast<__int128_t>((-9223372036854775807L-1)) - 1)
  Actual: "ffffffff"
Expected: "-8000000000000001"
../test/format-test.cc:1169: Failure
Value of: format("{0:x}", static_cast<__int128_t>((18446744073709551615UL)) + 1)
  Actual: "0"
Expected: "10000000000000000"
../test/format-test.cc:1171: Failure
Value of: format("{0:x}", int128_max)
  Actual: "ffffffff"
Expected: "7fffffffffffffffffffffffffffffff"
../test/format-test.cc:1172: Failure
Value of: format("{0:x}", int128_min)
  Actual: "0"
Expected: "-80000000000000000000000000000000"
../test/format-test.cc:1173: Failure
Value of: format("{0:x}", uint128_max)
  Actual: "ffffffff"
Expected: "ffffffffffffffffffffffffffffffff"
[  FAILED  ] FormatterTest.FormatHex (0 ms)
[ RUN      ] FormatterTest.FormatOct
../test/format-test.cc:1200: Failure
Value of: format("{0:o}", static_cast<__int128_t>((9223372036854775807L)) + 1)
  Actual: "0"
Expected: "1000000000000000000000"
../test/format-test.cc:1202: Failure
Value of: format("{0:o}", static_cast<__int128_t>((-9223372036854775807L-1)) - 1)
  Actual: "37777777777"
Expected: "-1000000000000000000001"
../test/format-test.cc:1204: Failure
Value of: format("{0:o}", static_cast<__int128_t>((18446744073709551615UL)) + 1)
  Actual: "0"
Expected: "2000000000000000000000"
../test/format-test.cc:1206: Failure
Value of: format("{0:o}", int128_max)
  Actual: "37777777777"
Expected: "1777777777777777777777777777777777777777777"
../test/format-test.cc:1208: Failure
Value of: format("{0:o}", int128_min)
  Actual: "0"
Expected: "-2000000000000000000000000000000000000000000"
../test/format-test.cc:1210: Failure
Value of: format("{0:o}", uint128_max)
  Actual: "37777777777"
Expected: "3777777777777777777777777777777777777777777"
[  FAILED  ] FormatterTest.FormatOct (0 ms)
[ RUN      ] FormatterTest.FormatIntLocale
[       OK ] FormatterTest.FormatIntLocale (0 ms)

[...]

[----------] Global test environment tear-down
[==========] 114 tests from 12 test cases ran. (14 ms total)
[  PASSED  ] 111 tests.
[  FAILED  ] 3 tests, listed below:
[  FAILED  ] FormatterTest.FormatDec
[  FAILED  ] FormatterTest.FormatHex
[  FAILED  ] FormatterTest.FormatOct

 3 FAILED TESTS

I think this is cause by the -DCMAKE_CXX_EXTENSIONS=OFF making the compilation command use -std=c++11 rather than -std=gnu++11.

[vitaut]

Looks like the same issue as #1725 which is already fixed in master.

[tonyelewis]

Ah 🤦‍♂️. Sorry - I didn't see #1725.

I don't think this is fully fixed in master, though there are now fewer failures. With the above instructions, I'm now seeing the following on master:

[...]

[ RUN      ] FormatterTest.FormatBin
[       OK ] FormatterTest.FormatBin (0 ms)
[ RUN      ] FormatterTest.FormatDec
../test/format-test.cc:1122: Failure
Value of: format("{0}", static_cast<__int128_t>((-9223372036854775807L-1)) - 1)
  Actual: "340282366920938463454151235394913435647"
Expected: "-9223372036854775809"
../test/format-test.cc:1128: Failure
Value of: format("{0}", int128_min)
  Actual: "170141183460469231731687303715884105728"
Expected: "-170141183460469231731687303715884105728"
[  FAILED  ] FormatterTest.FormatDec (0 ms)
[ RUN      ] FormatterTest.FormatHex
../test/format-test.cc:1163: Failure
Value of: format("{0:x}", static_cast<__int128_t>((-9223372036854775807L-1)) - 1)
  Actual: "ffffffffffffffff7fffffffffffffff"
Expected: "-8000000000000001"
../test/format-test.cc:1168: Failure
Value of: format("{0:x}", int128_min)
  Actual: "80000000000000000000000000000000"
Expected: "-80000000000000000000000000000000"
[  FAILED  ] FormatterTest.FormatHex (0 ms)
[ RUN      ] FormatterTest.FormatOct
../test/format-test.cc:1198: Failure
Value of: format("{0:o}", static_cast<__int128_t>((-9223372036854775807L-1)) - 1)
  Actual: "3777777777777777777776777777777777777777777"
Expected: "-1000000000000000000001"
../test/format-test.cc:1204: Failure
Value of: format("{0:o}", int128_min)
  Actual: "2000000000000000000000000000000000000000000"
Expected: "-2000000000000000000000000000000000000000000"
[  FAILED  ] FormatterTest.FormatOct (0 ms)
[ RUN      ] FormatterTest.FormatIntLocale
[       OK ] FormatterTest.FormatIntLocale (0 ms)

[...]

[----------] Global test environment tear-down
[==========] 114 tests from 12 test cases ran. (13 ms total)
[  PASSED  ] 111 tests.
[  FAILED  ] 3 tests, listed below:
[  FAILED  ] FormatterTest.FormatDec
[  FAILED  ] FormatterTest.FormatHex
[  FAILED  ] FormatterTest.FormatOct

[vitaut]

Should be fixed as part of #1725.

[tonyelewis]

I'm sorry to report that I'm still seeing the above test failures (with the failed tests' source line numbers increased by 1) in the current master (21c8b5c1421ae04df00a97e4c396ffb7b26916bc).

[vitaut]

Thanks for double-checking. There was one more issue with numeric_limits which is worked around in 51d0552.

[tonyelewis]

Yep - that seems to have done the trick. Thanks very much.

[tonyelewis]

I'm also seeing allocation-size-too-big from the AddressSanitizer but I'm assuming that's not worth reporting because it's from the test code with the comment "Check if std::allocator throws on allocating max size_t / 2 chars".