Had a real yak shave this morning with disabling assertions in release
builds in our C/C++ code. It turns out that our unit tests use
`assert()` which we intended to use from `test_assert.h`, but in some
cases `cassert` or `assert.h` had been #included after `test_assert.h`,
overriding our special `assert()` macro. The chain of includes is
somewhat hard to puzzle out -- it's often buried several levels deep.
This meant that a release build would drop all test assertions, meaning
most tests passed, unsurprisingly, as there were no assertions left to
fail ... but some tests failed with crashes because we optimized out
important lines such as `assert(some_important_function())`.
I was quite unhappy with this fragility, so I have opted to rename
`assert()` to `test_assert()` in all of our home-grown C/C++ unit tests,
which further highlighted unit tests which were only using the C/C++
`assert()` and not ours, so then had to figure out which unit test
executables needed to have `test_assert` added, and then ... then ...
discovered a bug in `test_color.h`, where we were #including
`io.h`/`unistd.h` inside a `namespace console_color {}` block, which
just happened to be the first ref to those beautiful headers, and thus
(because `#pragma once`) meant that useful little functions like
`access()` were no longer accessible to us in the global namespace.
I have also audited Every Single Call to `assert()` to verify that we do
not do Important Work inside the parentheses, and, apart from those
offending unit tests, now resolved with `test_assert()`, it looks like
all is good.
I would like to present one very well-shaved yak in this commit.
Fixes: #12619
- split keyboard loading into loading KMX file into blob and then
loading the keyboard processor from the blob.
- deprecate `km_core_keyboard_load`
- move file access next to deprecated method. This is now the only place
that loads a file in Core; unit tests have some more places that
load files.
- introduce GTest and add unit tests for loading from blob
Cherry-picked from `epic/web-core` branch.
Cherry-Pick-Commit: 1deaa323ad
Cherry-Pick-Commit: 59019cc8b7
Cherry-Pick-Commit: bc46458368
Cherry-Pick-Commit: d06aa29956
Cherry-Pick-Commit: 1c88166f6e
Cherry-Pick-Commit: 069cd21ecd
Cherry-Pick-Commit: 052ae2ec35
Cherry-Pick-Commit: 11a2a3ba3a
Part-of: #11293
Part-of: #8093
- temporary header file generated by wasm during build
- built using icu
- test_unicode verifies the contents and synchronization with running ICU.
- this is used by util_normalize to provide normalization properties under wasm without needing to include ICU.
Fixes: #9467
Most warnings have been cleaned up:
* min meson version is now 1.0
* additional parameters such as check, recursive added
* uses now global_source_root() instead of source_root()
* catting files on Windows wasn't actually working -- used in unit
tests. This is because meson passes paths with forward slashes to
cmd.exe `type` command, which doesn't understand them. This is bad,
because we were running effectively null tests for the affected tests.
Fortunately, the same tests were configured correctly on macOS and
Linux, and were all passing, so no serious damage.
There is one significant warning left:
`WARNING: Project targets '>=1.0' but uses feature deprecated since
'0.64.0': copy arg in configure_file. Use fs.copyfile instead`
Refer to mesonbuild/meson#12792. I have opened a PR against that to
undeprecate `copy` kwarg in a future version of meson.
Fixes: #8399
- load version data from node.js, Blocks.txt, and ICU4C
- support wasm: copy package.json, nodeversions.json and Blocks.txt into the keyboard area so that they can be mounted under wasm
also:
- rename 'fallback' macro to KMN_FALLBACK to not conflict with hedley in utfcodec.hpp
- fix ambiguous path type in tests
Fixes: #10183
Moves variable declaration for `bool enabled` from .h to .cpp and fixes
up builds. Fixes `#include` in ldml_test_source.cpp which was inside a
namespace.
Fixes: #10426
Adds the application context member of state and a corresponding,
failing unit test, which starts the process of verifying the handling
of the application context.
- lookup needs to handle either/neither/both cases for alt and ctrl
- and a bug in the test runner: would 'pass' if failed on exactly line 256 (0x100 & 0xFF == 0)
#7533
This updates our meson min version on Windows to 1.0, which means that
we can avoid having Visual Studio on the path, as meson can find it
anyway. With this change, we can eliminate the batch-wrapper for the
build, significantly improving build performance and simplifying the
scripts.
This also adds a change to disable tests for dependency builds of core,
along with the `--no-tests` command line option for doing builds without
tests.
Minor cleanup includes eliminating various warnings from meson.build
files, and standardizing WASM cross-platform build options for Windows.
- intermediate, does not process context or actions yet
- embedded (non-json) tests pass
- utf-8 parsing
- bring in kmx+ into ldml_test_source for parsing, but key2 is not ready eyt
- temporary scaffolding mapping key ids to vkeys
- build fixes
- getting closer.
The hextobin tool is used to help us construct invalid .kmx files
without having to either hack these into the compiler, or write them
with a hex editor. It feels a little bit of overkill, but this makes it
easy to (a) work with our intended file structures, and (b) review any
changes to these invalid keyboard tests.
Note: I did do a cursory search for a hex2bin or hextobin that we could
easily use but did not find a good simple cross-platform tool that we
could consume (`xxd -r` would work but AFAICT constrains our format to a
dump which is much harder to work with and read -- I wanted to be able
to annotate with file structures.)
For an example, see ik_000_null_invalid.txt.
This means that building the invalid keyboards for tests requires node,
but that should be fine given we need node to build the valid keyboards
anyway.
I've put hextobin under core/tools but it could just as well live under
common/tools in the future, if we find it useful for other project unit
tests.
Removes .kmx from repo and builds them from the source .xml files using
kmldmlc (building that if necessary also with a poor fella's dependency
check for now), and generally a lot of cleanup of the meson.build files
for the ldml tests.
Moves keyboards to a subfolder of the ldml unit test folder to keep them
neat and tidy and avoid embedding huge files into ldml.js for test of
WASM.
Note: 000_null_keyboard and 002_null_invalid do build successfully, but
the core ldml keyboardprocessor currently fails on them, because they
are missing keys and vkey sections.
- rewrite 000_null_keyboard.kmx to be valid, but with no keys
- remove outdated comment in kmx_plus.h
- update kmx_plus.cpp to not reverse the order of section ids (#7098)
- add a new test option for expected-error
