Squashed commit of the following: commit f36affd1a10943ab764a397e8a699d3f5194e9ae Author: Andy Wilson Date: Sat Mar 13 22:46:58 2021 -0700 Bug: Conda conflict between boost=1.74 and sphinx_rtd_theme This makes no sense, but there we are. I'm guessing that someone somewhere in the dependency chain pinned the wrong version. commit a79120df60962e7c9b33bab870df767b7b1ace61 Author: Andy Wilson Date: Sat Mar 13 22:16:14 2021 -0700 Bug: 'conda activate' fails with set -o nounused Some of the scripts that get run with 'conda activate' and 'conda deactivate' accidentally reference unused variables. This causes a failure when I'm running with strict debugging options like 'set -o nounused' in my script. While I was at it I added the -j/--parallel argument. The script now works for Python 3.9. It's throwing a conflict when creating the environment for Python 3.5 -- might need to special-case that one. commit 119711db4b838573758ba50ec71b6e9f4c47221b Author: Andy Wilson Date: Sat Mar 13 18:55:31 2021 -0700 Main loop now runs CMake We are now configuring the build! Next up: test (for debugging) whether the build tree works properly, then implement cmake_run_build, cmake_make_install, cmake_make_wheel, and copy_wheel_to_output. After that, dessert! commit 403ec1ea7f4397862875fb3b4041421b14faa88d Author: Andy Wilson Date: Sat Mar 13 18:00:20 2021 -0700 make_build_directory now uses BUILD_ROOT instead of an argument The container for all the build trees is a directory whose name is stored in BUILD_ROOT. Referring to this seems a little bit clearer than requiring it to be passed as an argument all the time. commit 6dcee12972337a3435970981a4b50b29b35774d7 Author: Andy Wilson Date: Sat Mar 13 17:44:31 2021 -0700 Refactor: reorder all functions except main() to be alphabetical The file should be a little easier to navigate now -- all the functions except main() and usage() are in alphabetical order. main() is last because, well, it's main(), and usage() is first because it's documentation. commit 7e70065d7e3addb2d792393625afbd03b958edd4 Author: Andy Wilson Date: Sat Mar 13 17:38:56 2021 -0700 Fix bugs in deleting BUILD_ROOT, activating Conda There is a difference between a variable being unset and a variable having the value "unset". I am now checking both when deleting BUILD_ROOT. Also, Anaconda's shell hooks have weird interactions with PS1 in a non-interactive shell. The workaround is 'export PS1=' (not a typo) before calling conda shell.bash hook. commit 860661929eb8c0c43244b87c7d7c75d83ed472a4 Author: Andy Wilson Date: Sat Mar 13 17:26:18 2021 -0700 Checkpoint: move main logic into main(), create Conda environments All of the logic is now in functions instead of immediate mode. I've wired up the main loop to start creating Conda environments. Next bug to fix: Conda's init hook needs to be run within the script, apparently? commit 1c783f6d940212a5cdb783926ecb39b124da3277 Author: Andy Wilson Date: Sat Mar 13 16:25:43 2021 -0700 Get my return semantics right (0 on success) Bash takes its cues from Unix command return values, which are always 0 for success and non-zero for anything else. Its tests do the same thing. A statement like if ! my_function; then do something fi translates to something like if (return_value_of(my_function) != 0) { do something } ...which is not what my accustomed-to-C brain was expecting. I believe I have fixed all of the functions so that they now return 0 on success. Wish me luck. commit a5ae0fa709d3a2471a8f38298ecec26a3338d2f6 Author: Andy Wilson Date: Sat Mar 13 15:06:25 2021 -0700 Fix bug in _conda_environment_exists I was using the syntax to retrieve an element from an associative array instead of the syntax to check whether an element in an associative array exists. As a programming language, Bash is weird and I don't like it. commit d61b074f783ff4939c0805ea911cfbbfbe71ddf8 Author: Andy Wilson Date: Fri Mar 12 22:18:15 2021 -0700 Infrastructure for 'conda create', main loop, caps for state variables There are now functions for checking to see that a Conda environment exists and creating one if necessary. I've also decided to name important variables in ALL_CAPS to distinguish between those and local no-big-deal variables. Finally, there is a main loop in place. All it does is echo the Python version it's building for and echo the steps it should take, but it's a step forward. Next up: add cleanup code to remove the build tree and sandbox directory. After that, flesh out the main loop one step at a time. commit 9e77a6586d74131959dc1e99769d5bc46b633456 Author: Andy Wilson Date: Fri Mar 12 19:09:00 2021 -0700 Change output semantics for all functions; make Conda environments a map Change #1: A function with name _foo now writes its output into variables named _foo_OUTPUT (if there's only one output) and _foo_OUTPUT_name if there are multiple outputs. Change #2: The function _list_conda_environments now returns its output as an associative array where the name of each virtualenv maps to its path. We'll need this in order to set BOOST_ROOT for CMake when we configure the build tree. commit c7542ac42ea9ab9e173cb6b66940f6b316d48564 Author: Andy Wilson Date: Thu Mar 11 21:33:45 2021 -0700 Add documentation to functions; add _list_conda_environments I'm including decent documentation for all of the functions in this script to facilitate reuse. In this checkpoint I added a _list_conda_environments function. This returns the list of Anaconda environments as an array. We will use it to check to see whether our build environments exist or need to be created. NOTE: need a pass through the Python versions array to remove duplicate elements commit b230643e0d2c6b1d524eb42289ff81664d71a318 Author: Andy Wilson Date: Thu Mar 11 20:12:51 2021 -0700 Add usage(), Python version syntax check, check for source dir The script now prints an error and a usage message if you get anything wrong. It also responds to -h / --help with the usage message. commit 78c9b648422df30e9296cb7f7a27d414fc7c983c Author: Andy Wilson Date: Thu Mar 11 19:28:05 2021 -0700 CHECKPOINT: working on build script for OSX wheels So far I'm just building infrastructure. Checking for prerequites, parsing command-line options. Next up will be checking the values of the Python versions to ensure that they meet proper syntax. There's a to-do list in the script. commit 50bf9f40cd0d869fdaecbeba766b4ff8d2848d4f Author: Andy Wilson Date: Mon Mar 8 20:36:11 2021 -0700 Update C++, CMake, Boost requirements I've bumped the required version to 1.74 (which will build with Python 3.9) and also told people to update CMake and their C++ compiler. Hopefully no-one will be left behind when we move to Boost 1.75. commit 46c11887b921762af9bc2a662826fb83bc0580de Author: Andy Wilson Date: Mon Mar 8 19:34:47 2021 -0700 Add missing boost::geometry #include There was a major reorganization of the internals of boost::geometry in Boost 1.75. Suddenly we were missing some important things because we included headers from inside the library instead of just the top-level . It might be worth checking to see whether we gain anything by trying to include only the most specific headers that we require. commit 4ccc20cf1b61f6439574643fdee31a2c25549fe4 Author: Andy Wilson Date: Sat Mar 6 02:36:32 2021 -0700 Fix deprecation warning about boost::bind placeholders Boost no longer defines the placeholders _1, _2, etc. in the global namespace. Instead, you must do one of the following things: 1. #include , then 'using namespace boost::placeholders' 2. #define BOOST_PLACEHOLDERS_GLOBAL_NAMESPACE (or some such) I'm going for choice #1. I'll change this if it turns out not to compile with older versions. commit d3b947d25b14d0cfedd5b93d9ebfec498f1b2d2a Author: Andy Wilson Date: Sat Mar 6 02:25:54 2021 -0700 Add missing include file for timestamp serialization Boost posix_time timestamps rely on the split_free function. In older versions, Boost would include the relevant header file (boost/serialization/split_free.hpp) itself. Now we have to do it. commit bff745309a36862ba93ec28a49d952413bb69e46 Author: Andy Wilson Date: Sat Mar 6 01:54:28 2021 -0700 Update CMakeLists to specify C++14 Rather than add a compile argument directly, we tell CMake to use C++14 so that it can add the relevant build-specific flag.

Add KML rendering to Python #1

Description

Activity

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions