- Boost - lists.preview.boost.org

[gil] Broken bit-aligned pixel locator on MSVC++ 64-bit optimized builds only
by Mateusz Loskot 14 Mar '20

14 Mar '20

Hi, Over the last few months, intermittently I've been chasing a peculiar bug that has been revealing itself in failing checksum tests in GIL's test/legacy/image.cpp [1] since we started testing GIL with MSVC++ (VS2017, VS2019) with b2 toolset=msvc variant=release address-model=64 [1] https://github.com/boostorg/gil/blob/2ff2d31895dae665ef3e05d0496bc082470e22… The peculiarity of this bug is that it has not been observed using any of these compilers: - GCC: 4.8, 4.9, 5, 6, 7, 8, 9 - Clang: 3.9, 4.0, 5.0, 6.0, 7, 8, 9, 10 - XCode: 8.3, 9.0, 9.1, 9.2, 9.3, 9.4, 10 There is complete report [2], but I also decided to reach the list looking for any insights that may help me to understand what may be causing the bug. Complete repo [3] has GitHub Actions to build and run minimal_test.cpp test. [2] https://github.com/boostorg/gil/issues/447 [3] https://github.com/mloskot/boost-gil-checksum-bug This minimal_test.cpp (at the bottom) leaves out the checksum calculation and narrows the scope of the bug down to the `xy_locator` and advances of the iterators. The test basically is this: 1. Create 3x3 image based on bit-aligned BGR or RGB pixel (e.g. bgr121_image_t) 2. Fill image with red 3. Draw blue diagonal ``` bgr121_image_t img(3, 3); // ... fill img with `bgr121_red` pixels // draw blue diagonal auto v = view(img); auto loc = v.xy_at(0, v.height() - 1); for (std::ptrdiff_t y = 0; y < v.height(); ++y) { *loc = bgr121_blue; // set blue pixel value ++loc.x(); // INCREMENT X FIRST --loc.y(); } ``` If the loop like this is compiled using either - `b2 toolset=msvc variant=release address-model=64 optimization=space` (/O1 /Ob2) - `b2 toolset=msvc variant=release address-model=64 optimization=speed` (/O2 /Ob2) then pixels of the blue diagonal are messed up. If the loop is compiled without optimization using either - `b2 toolset=msvc variant=debug address-model=64` (/Od /Ob2) - `b2 toolset=msvc variant=release address-model=64 optimization=off` (/Od /Ob2) then pixels of the blue diagonal are correct. Now, if the `for` loop above is modified, that is, lines incrementing the `x_iterator` and `y_itertor` are reordered: ``` for (std::ptrdiff_t y = 0; y < v.height(); ++y) { *loc = bgr121_blue; // set blue pixel value --loc.y(); ++loc.x(); // PRE-INCREMENT X SECOND } ``` or ``` for (std::ptrdiff_t y = 0; y < v.height(); ++y) { *loc = bgr121_blue; // set blue pixel value --loc.y(); ++loc.x(); // POST-INCREMENT X SECOND } ``` then the blue diagonal is drawn correctly regardless of the MSVC++ optimizations. If the loop is manually unrolled, see minimal_test.cpp below, the blue diagonal is correct too. I'm yet to try the CE to see compare the generated code, as far as my limited assembly fu goes, but currently Boost for MSVC++ on CE is broken (https://github.com/mattgodbolt/compiler-explorer/issues/1816) If anyone would have any insights on what may be causing the problem, or any suggestions what else to test, I would greatly appreciate that. Obviously, I'm getting suspicious about the MSVC++ optimizer, but I do know blaming a compiler is last+1 resort :) Thanks Mateusz ``` // minimal_test.cpp #include <boost/version.hpp> #include <boost/core/lightweight_test.hpp> #if BOOST_VERSION < 106900 #include <boost/gil/gil_all.hpp> #else #include <boost/gil.hpp> #endif namespace gil = boost::gil; #if BOOST_VERSION > 107200 using channel_sizes_t = boost::mp11::mp_list_c<int, 1, 2, 1>; #else using channel_sizes_t = boost::mpl::vector3_c<int, 1, 2, 1>; #endif using bgr121_ref_t = gil::bit_aligned_pixel_reference < std::uint8_t, channel_sizes_t, gil::bgr_layout_t, true > const; using bgr121_image_t = gil::image<bgr121_ref_t, false>; using bgr121_view_t = typename bgr121_image_t::view_t; using bgr121_pixel_t = typename bgr121_view_t::value_type; bgr121_pixel_t bgr121_red(0), bgr121_blue(0); void fill_image_red(bgr121_image_t& img) { gil::rgb8_pixel_t red8(255, 0, 0); gil::rgb8_pixel_t blue8(0, 0, 255); gil::color_convert(red8, bgr121_red); gil::color_convert(blue8, bgr121_blue); auto v = view(img); std::fill(v.begin(), v.end(), bgr121_red); for (auto it = view(img).begin().x(), end = view(img).end().x(); it != end; ++it) BOOST_TEST(*it == bgr121_red); } void test_draw_with_xy_locator_loop_fail(std::ptrdiff_t w, std::ptrdiff_t h) { bgr121_image_t img(w, h); fill_image_red(img); { auto v = view(img); auto loc = v.xy_at(0, v.height() - 1); for (std::ptrdiff_t y = 0; y < v.height(); ++y) { *loc = bgr121_blue; // OPTION 1 FAILS ++loc.x(); // PRE-INCREMENT X FIRST --loc.y(); // OPTION 2 FAILS TOO //loc.x()++; // POST-INCREMENT X FIRST //loc.y()--; // OPTION 3 FAILS TOO //auto& x_it = loc.x(); // PRE-INCREMENT NAMED X FIRST //++x_it; //auto& y_it = loc.y(); //--y_it; } } { auto it = view(img).begin().x(); // row 0 BOOST_TEST(*it == bgr121_red); ++it; BOOST_TEST(*it == bgr121_red); ++it; BOOST_TEST(*it == bgr121_blue); ++it; // row 1 BOOST_TEST(*it == bgr121_red); ++it; BOOST_TEST(*it == bgr121_blue); ++it; // FAIL BOOST_TEST(*it == bgr121_red); ++it; // row 2 BOOST_TEST(*it == bgr121_blue); ++it; BOOST_TEST(*it == bgr121_red); ++it; BOOST_TEST(*it == bgr121_red); } } void test_draw_with_xy_locator_loop_good(std::ptrdiff_t w, std::ptrdiff_t h) { bgr121_image_t img(w, h); fill_image_red(img); { auto v = view(img); auto loc = v.xy_at(0, v.height() - 1); for (std::ptrdiff_t y = 0; y < v.height(); ++y) { *loc = bgr121_blue; // OPTION 1 WORKS --loc.y(); ++loc.x(); // PRE-INCREMENT X SECOND // OPTION 2 WORKS TOO //loc.y()--; //loc.x()++; // POST-INCREMENT X SECOND } } { auto it = view(img).begin().x(); // row 0 BOOST_TEST(*it == bgr121_red); ++it; BOOST_TEST(*it == bgr121_red); ++it; BOOST_TEST(*it == bgr121_blue); ++it; // row 1 BOOST_TEST(*it == bgr121_red); ++it; BOOST_TEST(*it == bgr121_blue); ++it; BOOST_TEST(*it == bgr121_red); ++it; // row 2 BOOST_TEST(*it == bgr121_blue); ++it; BOOST_TEST(*it == bgr121_red); ++it; BOOST_TEST(*it == bgr121_red); } } void test_draw_with_xy_locator_step_good(std::ptrdiff_t w, std::ptrdiff_t h) { bgr121_image_t img(w, h); fill_image_red(img); { auto v = view(img); auto loc = v.xy_at(0, v.height() - 1); *loc = bgr121_blue; // red red blue ++loc.x(); --loc.y(); *loc = bgr121_blue; // red blue red ++loc.x(); --loc.y(); *loc = bgr121_blue; // blue red red ++loc.x(); --loc.y(); } { auto it = view(img).begin().x(); // row 0 BOOST_TEST(*it == bgr121_red); ++it; BOOST_TEST(*it == bgr121_red); ++it; BOOST_TEST(*it == bgr121_blue); ++it; // row 1 BOOST_TEST(*it == bgr121_red); ++it; BOOST_TEST(*it == bgr121_blue); ++it; BOOST_TEST(*it == bgr121_red); ++it; // row 2 BOOST_TEST(*it == bgr121_blue); ++it; BOOST_TEST(*it == bgr121_red); ++it; BOOST_TEST(*it == bgr121_red); } } int main() { test_draw_with_xy_locator_loop_fail(3, 3); test_draw_with_xy_locator_loop_good(3, 3); test_draw_with_xy_locator_step_good(3, 3); return boost::report_errors(); } ``` Best regards, -- Mateusz Loskot, http://mateusz.loskot.net

1 1

[website] Update notes of historic releases
by Mateusz Loskot 11 Mar '20

11 Mar '20

Hi, Is this possible to update release notes of historic releases? I'd like to update notes of GIL for 1.69 which, as I found, have been never collected and are completely missing https://github.com/boostorg/gil/commit/2ff2d31895dae665ef3e05d0496bc082470e… Best regards, -- Mateusz Loskot, http://mateusz.loskot.net

2 2

C++ Deflate (zlib-like) library
by Janson R. 11 Mar '20

11 Mar '20

Hello everyone, I have recently been working on a C++ compression library very similar to zlib after trying to implement some HTTP compression support over Boost.Beast and realizing after some discussion with sir Falco that while it would be a nice builtin feature for Beast, it would possibly be a better idea to have zlib-like compression be a separate library in order to be properly maintainable and likely more useful. The current working version can be viewed at https://github.com/ryanjanson/Deflate however the API could still be helped in terms of modern C++ usability, and what I currently have in mind can be found annotated here: https://gist.github.com/AeroStun/687ec9ca69404e26f8e02e5084926036 Do you thing that this could be useful to have as its own entity in the Boost environment? Any kind of feedback on the idea and the library is warmly welcome. Regards, Janson

14 58

boost thread strange behaviour
by Adham Sabry 08 Mar '20

08 Mar '20

hi , i am trying to learn executors and executors thread pools using with boost::async but i found strange behabiour when i make thread pool adaptor ea and submit tasks to it and make ea00 and submit tasks to it if i started ea with async , it automatically starts ea00 the attached file has the code i thought that async will only start the passed thread pool"adaptor" But i think there is underlying thing which gather all threads of different pools into single entity and starts them with any call to boost::async is this a bug??? i run this from windows 7 64 by visual studio 2017

3 7

[GSoC20] [Boost.Geometry] Seeking mentor for project 1
by João Tavares 07 Mar '20

07 Mar '20

Hello, I am a third year undergraduate student from IST - University of Lisbon in Applied Mathematics and Computer Science. I am very interested in the project 1 on the Boost.Geometry library. Having dealt with this sort of problem both on my university courses and on my own, in the context of competitive programming, I feel like I could contribute to this project during GSoC 2020. Regarding the competency test I have not been able to understand how or where it should be submitted. I am sending this e-mail to know whether there is someone willing to mentor me for this project. Thanks for the replies in advance. Best regards, João Tavares IST - University of Lisbon

2 1

ublas GSOC'20 Project
by Mrityunjay Tripathi 07 Mar '20

07 Mar '20

Hi everyone, I am Mrityunjay Tripathi, a 3rd-year Computer Science & Engineering undergraduate. I have been getting familiar with the boostorg/math and boostorg/ublas repositories for quite a few weeks. I find it quite interesting in terms of the functionalities they provide. But I see that the tensors section of uBLAS can be more improved by parallelizing the code using OpenMP and providing support for cuBLAS. I also wanted to add flexible NumPy/PyTorch tensor features such as autograd on tensors. Any feedback or suggestions are most welcome. Best Regards, Mrityunjay Tripathi

1 0

Hello!
by Kishan shukla 07 Mar '20

07 Mar '20

Hello Everyone! My name is Kishan Shukla, and I am a third-year undergraduate student of mathematics at IIT Kanpur. I am interested in contributing to Boost.real and have already sent two PR's https://github.com/BoostGSoC19/Real/pull/40 and https://github.com/BoostGSoC19/Real/pull/41, which are yet to be reviewed. Currently, I am stuck at a problem that I want to discuss with the mentors. Can I ask the project-specific issues here, or are there project-specific channels?

2 1

[build] ERROR: rule "[[bbv2.builtin.features.warnings]]`warnings`::" unknown in module "warnings-feature"
by Mateusz Loskot 06 Mar '20

06 Mar '20

Hi, (CC'ed to boost and boost-build) Two of Boost.Geometry users reported that peculiar error, [1] and [2] ``` /home/digu_j/boost/tools/build/src/tools/features/warnings-feature.jam:27: in load ERROR: rule "[[bbv2.builtin.features.warnings]]`warnings`::" unknown in module "warnings-feature". ``` and I'm having a hard time understanding what is going wrong here and what causes this issue, on Linux. The line 27 is a typical Jam line: ``` feature.feature warnings : on all extra pedantic off : incidental propagated ; ``` but this refers to something that puzzles me most ``` ERROR: rule "[[bbv2.builtin.features.warnings]]`warnings`::" unknown ``` a bit of comment block? ``` #| tag::doc[] [[bbv2.builtin.features.warnings]]`warnings`:: ``` Is this a parsing issue? Is this related to potentially existing system-wide installation of Boost.Build? If yes, why none of the dumps in [1] and [2] refer to any files in /usr? [1] ./b2 headers attempt https://github.com/boostorg/geometry/pull/667#issuecomment-584794370 [2] b2 attempt to build a program in libs/geometry https://github.com/boostorg/geometry/pull/665#issuecomment-595595900 Best regards, -- Mateusz Loskot, http://mateusz.loskot.net

2 2

[date-time] USE_DATE_TIME_PRE_1_33_FACET_IO to be removed in next release
by Jeff Garland 06 Mar '20

06 Mar '20

//crosspost from users list In case you've not checked the calendar (see below) it's 2020 and a the c++ world has moved on from the issue this macro solved. boost date-time also needs to move forward into the decade with a modernized all header code base . If you're dependent on this macro for any reason I'd like to hear about it. Thanks, Jeff using namespace boost::gregorian; date today = day_clock::local_day(); cout << today.year() << endl; //2020 ;)

1 0

Help: [Boost.GIL] setup
by R Rajiv 06 Mar '20

06 Mar '20

Hi I am Rajiv, a student whose looking to contribute to Boost's GIL library this year for GSOC 2020. I have read through the contribution.md page for setup of the environment for Boost.gil @ https://github.com/boostorg/gil/blob/develop/CONTRIBUTING.md. As I do not have cloning rights for the repo boost.git, I have forked and cloned my fork with --recurse-submodules.This leads me to the following issue: _______ fatal: repository 'https://github.com/rRajivramachandran/accumulators.git/' not found fatal: clone of 'https://github.com/rRajivramachandran/accumulators.git' into submodule path 'libs/accumulators' failed ________ I have been unable to proceed due to this issue. -- Thanks Rajiv

3 2