Fork of https://github.com/zephyrproject-rtos/Kconfiglib.git https://git.mcdonnell.dev/Kconfiglib.git/atom?h=v1.0.4 2017-09-29T19:49:50Z 2017-09-29T19:49:50Z Ulf Magnusson ulfalizer@gmail.com 2017-09-29T19:49:50Z urn:sha1:c8f8e742fce13c0cd86f1857785de141549aa781 Fixes: - After 250b370 (Refactor load_config() and fix minor issues), print_undef_assign=True also considers "# CONFIG_NOT_DEFINED is not set", which acts like an "n" assignment in practice. - 7cbfa47 (Fix undef. assign warning in set_user_value()) Also contains some internal cleanup and optimization. 2017-09-29T18:07:39Z Ulf Magnusson ulfalizer@gmail.com 2017-09-29T15:30:34Z urn:sha1:250b370cb5a1d40d7c0e2a4df7275964c1d24fea Minor fixes: - Even with print_undef_assign=True, "# CONFIG_NOT_DEFINED is not set" would not generate a warning. - Change the warning when a variable is set more than once in the .config to be a bit more explicit. - Only de-quote and unescape assignments to string variables, which is nicer and matches the C implementation. 2017-09-29T14:45:47Z Ulf Magnusson ulfalizer@gmail.com 2017-09-29T14:26:05Z urn:sha1:c54db8e7c9cde50a2b13cebe256cfd907b45d762 Rearrange a bit and document that a trailing newline does not need to be added. Clean up the tests a bit too. 2017-09-29T12:48:54Z Ulf Magnusson ulfalizer@gmail.com 2017-09-29T11:56:20Z urn:sha1:7cbfa4711033e7f0d6b730a2ad22fd1caa84beea Previously we'd crash instead of printing the warning, because we didn't pass 'filename' and 'linenr' to _stderr_msg(). This code path currently can't be reached via load_config() and needs an explicit set_user_value(), so the bug went undetected. Fix by breaking out a separate _warn_undef_assign() that defaults 'filename' and 'linenr' to None, similar to _warn(). 2017-09-29T04:15:01Z Ulf Magnusson ulfalizer@gmail.com 2017-09-29T01:49:35Z urn:sha1:5fea36ffd71d8f908133d091860ef4b139cc7e81 By consistently stripping trailing whitespace instead of stripping initial whitespace in the string/operator case, we do less redundant work and handle the '\n's at the end of lines better. Shaves a few % off the _tokenize() runtime in cProfile and line_profiler. 2017-09-29T04:14:39Z Ulf Magnusson ulfalizer@gmail.com 2017-09-29T00:19:20Z urn:sha1:3d27a14dbcb057eb99491babe1238bde6cfe66e6 More explicit. 2017-09-29T04:03:38Z Ulf Magnusson ulfalizer@gmail.com 2017-09-29T00:10:55Z urn:sha1:35e8acdb2d49f279862a99ea129a9bb260b0bc4a Think I had it in the back of my head somewhere that not invalidating undefined symbols could break some obscure cases, but turns out it's perfectly safe: Nothing can change the value of an undefined symbol. They always get their name as their value. There's no need to unset user values on them either, because set_user_value() already refuses to to set one on them. Lets us get rid of the Python 2/3 compatibility hack and instead iterate over a plain list of defined symbols. 2017-09-29T03:55:02Z Ulf Magnusson ulfalizer@gmail.com 2017-09-28T23:01:02Z urn:sha1:f56ebf74f470431b0d262d2ce6f95d442b58f2f9 The clear loop in write_config() already guarantees that _already_written will be set for all symbols that could be written out to the .config. Clarifies the scope. Also change the code to only clear _already_written for defined symbols. No other symbols could ever be written out. Gives a nice loop over a plain list. 2017-09-29T03:53:17Z Ulf Magnusson ulfalizer@gmail.com 2017-09-28T22:31:56Z urn:sha1:e7c6c7a5f5ca20844cf23655a4a37fa7ccd6fc8f 2017-09-28T18:49:50Z Ulf Magnusson ulfalizer@gmail.com 2017-09-28T18:47:25Z urn:sha1:30cdddbdb48ffa0562d740bb7d0d78aedc98a96c Would get crammed in between the arch/defconfig string and the "FAIL".