Gitiles
Code Review Sign In
gerrit-public.fairphone.software / platform / external / google-fruit / ef4bc1fddc438276708917e2341b80c37a76100f
  1. ef4bc1f Replace tests that have a non-annotated and an annotated variant with a single parameterized test (per pair). by Marco Poletti · 8 years ago
  2. e7ac9ba Split some tests that were testing multiple things. by Marco Poletti · 8 years ago
  3. 824579c Add an empty main() to the test source code if it doesn't have a main, to make some tests more concise. Also omit the explicit "return 0" in main, since that's added implicitly by the compiler anyway. by Marco Poletti · 8 years ago
  4. c993c12 Split a test into two more focused ones. by Marco Poletti · 8 years ago
  5. 8dbb51a Add a Markdown document with the result of a (manual) assessment of the e2e test coverage, in preparation for adding more tests. by Marco Poletti · 8 years ago
  6. 6df80a7 Use extras/bazel_root as WORKSPACE root for Bazel, using symlinks so that Fruit lives under //third_party/fruit instead of //. by Marco Poletti · 8 years ago
  7. 8072dca Re-add support for running e2e Fruit tests with bazel. by Marco Poletti · 8 years ago
  8. d873e2f Don't set -pedantic with GCC 4.8, it's too restrictive (e.g. it doesn't allow casting a function pointer to void*). by Marco Poletti · 8 years ago
  9. e9997dd Add Clang 3.9 to the dockerfiles used for CI testing. by Marco Poletti · 8 years ago
  10. af3ff96 Revert to using Clang 3.8 in OS X tests, the Clang 3.9 tests are failing due to an installation issue. by Marco Poletti · 8 years ago
  11. e63ebc7 Disable the missing-braces warning in release mode too. by Marco Poletti · 8 years ago
  12. 51ff15b Stop testing with ASan/UBSan in old Ubuntu versions (in Travis CI). They're unlikely to uncover real problems and they increase the size of the build matrix. by Marco Poletti · 8 years ago
  13. 218458f Test with Clang 3.9 and GCC 6 on OS X, instead of Clang 3.8 and GCC 5. by Marco Poletti · 8 years ago
  14. cd7f0b3 Disable the missing-braces error in tests, it's too noisy (see also https://llvm.org/bugs/show_bug.cgi?id=21629). GCC no longer has this in Wall, but Clang still does. by Marco Poletti · 8 years ago
  15. 53f136c Remove more using declarations from tests. by Marco Poletti · 8 years ago
  16. 8b2abe6 Compile with -pedantic in debug mode, and fix a few issues that were reported, notably a declaration of a 0-length array that isn't standard C++ (even though GCC and Clang accept it). Fixes issue 26. by Marco Poletti · 8 years ago
  17. 6317c25 Always use 'struct Foo' instead of 'class Foo' in tests. Clang complains if these are mixed. by Marco Poletti · 8 years ago
  18. 83b1ad8 Use explicit fruit:: qualifications in tests, instead of relying on using declarations. This is to avoid hiding real bugs. by Marco Poletti · 8 years ago
  19. 1c7e821 Make the FruitAssistedTypedef and FruitAnnotatedTypedef typedefs public, so that the INJECT macro no longer needs to contain private:/public: specifiers. The public: specifier could have unintentionally exposed private parts of a class if INJECT is used in the private part of a class. by Marco Poletti · 8 years ago
  20. 21111b0 Make e2e tests more concise and readable by factoring common declarations/definitions at the beginning of each test class. by Marco Poletti · 8 years ago
  21. cb9e584 Fix the bytes->str conversion in the e2e testing code (again). by Marco Poletti · 8 years ago
  22. fe8aa52 No longer test on OS X with clang 3.7 (since we're testing with 3.6 and 3.8) and XCode 7.3 (since we're testing with 7.1 and 8). by Marco Poletti · 8 years ago
  23. 3b4a2ed Run some quick 'smoke tests' at the beginning of the test run in Travis CI, to find most issues sooner. by Marco Poletti · 8 years ago
  24. a1ebbf2 Don't run with ASan/UBSan on OSX with Clang 3.8, it doesn't work. by Marco Poletti · 8 years ago
  25. 7522a33 Disable ASan/UBSan in combinations where they don't work (in Travis CI). by Marco Poletti · 8 years ago
  26. a816ce4 Simplify postsubmit.sh, removing bits that are no longer needed now that there's a single postsubmit.sh run per row in the Travis CI build matrix. by Marco Poletti · 8 years ago
  27. fdb4155 Check in a simple git hook to check that the .travis.yml file is updated. by Marco Poletti · 8 years ago
  28. 92b2a02 Add a comment at the top of .travis.yml to highlight the fact that it was auto-generated. by Marco Poletti · 8 years ago
  29. 0d8c181 Re-add comments on disabled tests that used to be in .travis.yml into extras/scripts/travis_yml_generator.py. Also re-enable a test that was disabled by mistake. by Marco Poletti · 8 years ago
  30. 297a6bb Export the OS env variable in .travis.ci too, the presubmit script expects it. by Marco Poletti · 8 years ago
  31. 116e1da Various changes to the .travis.yml generator: properly export environment variables, run additional tests that were previously commented-out, and make the test name appear in the Travis CI dashboard. by Marco Poletti · 8 years ago
  32. 6938782 Dynamically-generate the .travis.yml file, for increased readability and conciseness. Also split rows of the test matrix to increase parallelization where possible. by Marco Poletti · 8 years ago
  33. 55ec5fd Fix a build issue introduced by the pull request #33. by Marco Poletti · 8 years ago
  34. 611a02e Merge pull request #33 from cygnus9/master by poletti-marco · 8 years ago
  35. 6b4ad7f Use a more general Valgrind suppression rule in tests (hoping to fix GCC tests on OSX, that are failing ATM). by Marco Poletti · 8 years ago
  36. 283873e Add even more Valgrind suppression rules (needed for OS X tests), the Travis CI tests are failing again, probably due to an update of the OS X system they use to test. by Marco Poletti · 8 years ago
  37. 8b63071 Always print stdout/stderr as strings in e2e tests (when they fail), instead of occasionally printing them as bytes. by Marco Poletti · 8 years ago
  38. 36efab5 Print the entire stderr/stdout in nose2-based e2e tests (except the ones where a compiler error is expected). by Marco Poletti · 8 years ago
  39. 8105f6b Make the unittests actually compile the tests when fruit is used as a submodule by Mark Swaanenburg · 8 years ago
  40. eb724f3 Make project usable as submodule by Mark Swaanenburg · 8 years ago
  41. bc45f7d Disable ASan on OSX with GCC 4.8, it doesn't work anymore due to a bug in glibc by poletti-marco · 8 years ago
  42. e6e9045 Install Clang 3.8 on OSX before testing with it in Travis CI. by Marco Poletti · 8 years ago
  43. ac2007b Add a couple more Valgrind error suppressions, for errors currently happening in OS X (that have nothing to do with Fruit). by Marco Poletti · 8 years ago
  44. cad06d5 Add Clang 3.8 on OS X to the Travis CI build matrix. by Marco Poletti · 8 years ago
  45. 4cf5205 Fix various test failures due to the python test rewrite. by Marco Poletti · 8 years ago
  46. 5498f4f Fix testing dockerfiles for 14.04 and 15.10: the python3-sh package didn't exist at that time, install it via pip instead. by Marco Poletti · 8 years ago
  47. 530c4c0 Various improvements/fixes to the nose-based tests. Notably, adding nose2 to the basesystem dockerfiles used in Travis CI and moved from nose2's multiprocess to running each python file sequentially and relying on ctest's parallelization. by Marco Poletti · 8 years ago
  48. f9b2c6f Port all end-to-end C++ tests to be nose2-based python tests, so that: by Marco Poletti · 8 years ago
  49. 2d34eac Add an internal YML file with tables that have more detail than the ones in the wiki. by Marco Poletti · 8 years ago
  50. d753167 Add some synthetic dimensions to benchmark results: di_library_git_commit_hash and di_library_version_name, in addition to compiler_name which no longer has special handling. by Marco Poletti · 8 years ago
  51. c75d608 Fix typo in doxygen comment. by Marco Poletti · 8 years ago
  52. e4db0ab Add a simple benchmark definition yml file with a single benchmark. by Marco Poletti · 8 years ago
  53. b7fe050 Fix the benchmarking code so that it works with both Fruit <2.1.0 and >=2.1.0. by Marco Poletti · 8 years ago
  54. 5aed85a Add a --continue-benchmarks option to run_benchmarks.py. by Marco Poletti · 8 years ago
  55. 3c07b9f Increase the number of iterations in each loop of the runtime benchmarks by 2.5x, there's too much variance otherwise. by Marco Poletti · 8 years ago
  56. 1559d1f Split the yml file for Fruit and Boost.DI benchmarks so that benchmarks can be easily repeated for new versions of each. by Marco Poletti · 8 years ago
  57. 584a6cd Minor simplification to the benchmark code, using defaultdicts instead of setdefault(). by Marco Poletti · 8 years ago
  58. 8223fe5 Execute lazyGetPtr() calls all together before the corresponding getPtr() calls. This improves per-request performance by ~10%, but causes an increase of ~15-20% of the executable size. by Marco Poletti · 8 years ago
  59. 04cf924 Various minor executable size optimizations. by Marco Poletti · 8 years ago
  60. 6414851 Re-inline some component_storage methods, the increase in object size is relatively small. by Marco Poletti · 8 years ago
  61. 0316a00 Reimplement the storage for PartialComponent to: by Marco Poletti · 8 years ago
  62. 677df8d Add all top-level build* directories to gitignore. by Marco Poletti · 8 years ago
  63. 26cf514 Add a simple script to find Fruit implementation types without unit tests. by Marco Poletti · 8 years ago
  64. cf5c9f1 No longer inline some component-related code, to reduce the executable size of client code using Fruit. This yields a ~6x reduction in the benchmark with 100 classes (1.6MB -> 275KB). by Marco Poletti · 8 years ago
  65. 2b13ff9 Add a script to analyze the symbol size in the generated benchmark code. by Marco Poletti · 8 years ago
  66. b22aceb Use rpath flags for dynamic linking of benchmark code, instead of using LD_LIBRARY_PATH. by Marco Poletti · 8 years ago
  67. feaed5c Report an error when passing a pointer to bindInstance(), instead of reporting an obscure error later on. by Marco Poletti · 8 years ago
  68. dcd6e6e Remove obsolete targets from the BUILD file for benchmarks. by Marco Poletti · 8 years ago
  69. d94b25e Fix a couple remaining issues with the benchmark code and introduce YAML files to configure the tests to run and the tables to produce with the test results. Also improve code quality of benchmarking code. by Marco Poletti · 8 years ago
  70. 1a23d0b Introduce python scripts to run multiple benchmarks and analyze the results. by Marco Poletti · 8 years ago
  71. 8960354 Fix build error, by removing a reference to a no-longer-existing dir. by Marco Poletti · 8 years ago
  72. bc1c7bc Rewrite the benchmark code in python, slightly changing the generated benchmark, and add a Boost.DI benchmark generator. by Marco Poletti · 8 years ago
  73. bf10015 Correctly set ASAN_OPTIONS in tests as desired. by Marco Poletti · 8 years ago
  74. 549e7d4 Disable malloc vs new consistency checks in Asan when running tests on OS X using GCC 4.8. by Marco Poletti · 8 years ago
  75. 94b5a95 Ignore some Valgrind-reported leaks in tlv_get_addr on OS X, as they are not real issues in Fruit. by Marco Poletti · 8 years ago
  76. e32edc2 Add more Valgrind suppressions, to suppress some reported errors in __emutls_get_address that happen on OS X. by Marco Poletti · 8 years ago
  77. 1789606 Remove a source file that was committed by mistake. This was causing build failures with Bazel (since it uses globs). by Marco Poletti · 8 years ago
  78. d07bd8a Fix an 'unused variable' warning reported by Clang (that was breaking the build with -W -Wall). by Marco Poletti · 8 years ago
  79. ba4d179 No longer include boost::hash{set,map} in fruit.h (not even as forward decls). by Marco Poletti · 8 years ago
  80. 3e576c2 No longer include normalized_component_storage.h in fruit.h. by Marco Poletti · 8 years ago
  81. 62f2574 Reduce the amount of code included when including fruit.h. by Marco Poletti · 8 years ago
  82. c6d5955 Use a singly-linked list instead of double-linked. This improves performance. by Marco Poletti · 8 years ago
  83. 29c8cf4 Save memory and slightly improve performance by storing a BindingData in 3*sizeof(void*) bytes instead of 4*sizeof(void*). by Marco Poletti · 8 years ago
  84. 6bbf6c9 Remove the method SemistaticGraph::changeNodeToTerminal(), it was never used. by Marco Poletti · 8 years ago
  85. 632eff4 Ignore all bazel-* files/directories in .gitignore. by Marco Poletti · 8 years ago
  86. efb74c1 Start testing against XCode 8 (ATM in beta) in Travis CI. by Marco Poletti · 8 years ago
  87. 3009aac Re-enable tests against Clang on OS X installed via Brew, and disable the ones using XCode. Those are the ones that don't work. by Marco Poletti · 8 years ago
  88. b537848 Stop supporting Clang installed via Brew on OS X. It doesn't support thread_local. by Marco Poletti · 8 years ago
  89. c1e07e4 Remove ${} from arguments to a CMake if to avoid expansion problems. by Marco Poletti · 8 years ago
  90. 7542a33 Don't link with libsupc++ on OS X, it causes a "library not found" error. by Marco Poletti · 8 years ago
  91. 78e1e53 Add an explicit "-lm" linker flag when building with Bazel, this seems to be needed now for some reason (and it's currently causing the build failure with Bazel in Travis CI). by Marco Poletti · 8 years ago
  92. 205f9df Explicitly link against libsupc++. It's needed now when using Clang and libc++ (not sure why it wasn't before), and it's not harmful when using libstdc++. by Marco Poletti · 8 years ago
  93. e7a6d4c Remove a superfluous std::move() that was actually decreasing performance (following a Clang warning). by Marco Poletti · 8 years ago
  94. 8db9216 Use lists with a custom allocator instead of caching vectors. by Marco Poletti · 8 years ago
  95. 9722c4d Optimize component construction (before normalization) by caching binding vectors instead of repeatedly destroying and reconstructing them. by Marco Poletti · 8 years ago
  96. 89baa7e Reduce build verbosity in Travis CI. by Marco Poletti · 8 years ago
  97. 3d6a23f Add boost to the docker images used for CI testing of Fruit. by Marco Poletti · 8 years ago
  98. 4297358 Add boost as a build dependency in the packaging scripts. by Marco Poletti · 8 years ago
  99. 886289c Switch back to an STL hashmap for multibindings and debug-only maps, and avoid including Boost's hashmap in public headers to reduce the compile time of code using Fruit. by Marco Poletti · 8 years ago
  100. d4eda1f Set basic release flags when Fruit is compiled with CMAKE_BUILD_TYPE=Release (or no explicit CMAKE_BUILD_TYPE), to prevent people from compiling in non-release mode or with assertions enabled by mistake. by Marco Poletti · 8 years ago