Builds with -DNDEBUG do not contain the compilation string, so the test
will fail even if nvim is built with unibilium.
Similarly, zig builds only report "Compilation: TODO", so they also fail
the test even when build with unibilium
Detect both scenarios and skip the test, rather than failing it.
PR #34876 expanded the total range of values that cjson considers
valid. However, it didn't address the bigger problem of storing a
`long long` value in a `lua_Integer` (which is typically a typedef for
`ptrdiff_t`).
On 32-bit platforms, this ends up storing an 8-byte number into a 4-byte
variable, truncating the value.
Store the converted value in a temporary `long long` variable so we can
detect the scenario and decode into a `lua_Number`.
As noted in #34908, the arm unittests fail due to unrecognized types:
ERROR test/unit/testutil.lua @ 802: Expressions parser works with &opt
test/unit/testutil.lua:774: test/unit/testutil.lua:758: (string) '
test/unit/testutil.lua:288: declaration specifier expected near '__SVFloat32_t''
exit code: 256
After testing on Debian's arm64 porterbox, the unittests cleanly pass
when ignoring various types defined in
/usr/include/aarch64-linux-gnu/bits/math-vector.h
A binary format spec always expects a corresponding unsigned long long
value. However, that explicit handling didn't get included when porting
the code from Vim, so binary format spec was falling through to the
"unsigned" and "length_modifier = NUL" portion of the code:
} else {
// unsigned
switch (length_modifier) {
case NUL:
uarg = (tvs
? (unsigned)tv_nr(tvs, &arg_idx)
: (skip_to_arg(ap_types, ap_start, &ap, &arg_idx,
&arg_cur, fmt),
va_arg(ap, unsigned)));
break;
This incorrectly read an "unsigned" value from an "unsigned long long"
variable, which would produce incorrect results on certain platforms.
Same idea as a7be4b7bf8, but that only showed the context if the
length of the string differed. Since these tests check both string
length and string content, the ctx should be provided for both.
ERROR test/unit/testutil.lua @ 797: vim_snprintf() positional arguments
test/unit/testutil.lua:769: test/unit/testutil.lua:753: (string) '
test/unit/strings_spec.lua:159: snprintf(buf, 4, "%1$0.*2$b", 12ULL, cdata<int>: 0xf78c8ed8) = 001100
Expected objects to be the same.
Passed in:
(string) '000'
Expected:
(string) '001''
Since these assertions all use a common function to perform the test
assertions, it's difficult to figure out which test failed:
ERROR test/unit/testutil.lua @ 785: vim_snprintf() positional arguments
test/unit/testutil.lua:757: test/unit/testutil.lua:741: (string) '
test/unit/strings_spec.lua:143: Expected objects to be the same.
Passed in:
(number) 6400
Expected:
(number) 6'
exit code: 256
Adding context to the assertion makes it clearer what the problem is:
ERROR test/unit/testutil.lua @ 785: vim_snprintf() positional arguments
test/unit/testutil.lua:757: test/unit/testutil.lua:741: (string) '
test/unit/strings_spec.lua:149: snprintf(buf, 0, "%1$0.*2$b", cdata<unsigned int>: 0xf78d0f38, cdata<int>: 0xf78dc4e0) = 001100
Expected objects to be the same.
Passed in:
(number) 6400
Expected:
(number) 6'
exit code: 256
In order to run the marktree unit test in release mode, the test functions need to be available even when NDEBUG is defined.
Keep the body of marktree_check a nop during release builds, which limits the usefulness of the testing, but at least lets the tests run.
In release builds, the Compilation: line is omitted so the build is reproducible. Since the "fall-back for $VIM" line is always present, check for that instead.
Tests should not rely on being run inside a git clone, so the Linux
builds cover this use case. The macOS builds will continue running with
the .git directory so there's still unix-ish coverage within a git
clone.
This ensures that no tests fail due to differences between release and debug builds.
The release build-type check is now unnecessary, too, so remove it.
In order to run unittests with a release build, we need the test
functions to be accessible when NDEBUG is defined. Moving the functions
into the test fixture ensures they are available and only available for
use by the unit tests.
`get_target_property(<var> ...)` sets `<var>` to `<var>-NOTFOUND` if the
property doesn't exist for the given target. Detect this situation to
avoid adding various `-Dprop-NOTFOUND` and `-Iprop-NOTFOUND` to the
command-line when generating the headers.
buf is a pointer argument, not a local char array, so sizeof(buf) is just the size of a pointer type on the platform. This is always an incorrect value, but on 32-bit platforms it actually has an impact, since sizeof(buf) is just 4 and causes the buffer to get truncated.
Since execl() is a variadic function, it requries a NULL-terminal to
indicate the end of its argument list, c.f. exec(3)
> The first argument, by convention, should point to the filename
> associated with the file being executed. The list of arguments *must*
> be terminated by a null pointer
This fixes the failure seen on aarch64 and i386, due to garbage data
being considered part of the variadic arguments.
16 bytes is not enough room for existing usage of the buffer by the
tests, so the name may get truncated and cause log_spec test to fail:
FAILED test/functional/core/log_spec.lua @ 30: log messages are formatted with name or test id
test/helpers.lua:146: retry() attempts: 51
test/helpers.lua:155: Pattern "%.%d+%.%d/c +server_init:%d+: test log message" not found in log (last 100 lines): Xtest_logging:
ERR 2023-11-24T23:36:34.252 T1274.2445945.0 server_init:57: test log message
ERR 2023-11-24T23:36:34.275 T1274.2445945.0 server_init:57: test log message
09d4133 changed blocknr_T from long to int64_t, so pe_bnum is now always 64-bit. This was an incompatible change in the swapfile format for 32-bit systems, but there have been no complaints in the past 9 years so just adjust the test.
Now that all the Find<Foo>.cmake modules are using our custom
find_path2, the bundled deps build needs to include the Find module so
it has access to find_path2.
The missing include was causing the PPA build to fail:
```
cd .deps && \
cmake -G 'Unix Makefiles' \
-DUSE_EXISTING_SRC_DIR=ON -DUSE_BUNDLED_LUAROCKS=OFF -DUSE_BUNDLED_LUAJIT=OFF -DUSE_BUNDLED_GPERF=OFF -DUSE_BUNDLED_LUA=OFF /<<PKGBUILDDIR>>//cmake.deps
-- The C compiler identification is GNU 7.5.0
-- Check for working C compiler: /usr/bin/cc
-- Check for working C compiler: /usr/bin/cc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Found GNU Make at /usr/bin/make
-- CMAKE_BUILD_TYPE not specified, default is 'Debug'
-- Performing Test HAS_OG_FLAG
-- Performing Test HAS_OG_FLAG - Success
-- Found Git: /usr/bin/git (found version "2.17.1")
CMake Error at /<<PKGBUILDDIR>>/cmake/FindLuajit.cmake:1 (find_path2):
Unknown CMake command "find_path2".
Call Stack (most recent call first):
cmake/BuildLuv.cmake:13 (find_package)
CMakeLists.txt:127 (include)
-- Configuring incomplete, errors occurred!
```
Linux added these types to their userspace headers in [6.5], which
causes unit tests to fail like
```
-------- Running tests from test/unit/api/private_helpers_spec.lua
RUN vim_to_object converts true: 17.00 ms ERR
test/unit/helpers.lua:748: test/unit/helpers.lua:732: (string) '
test/unit/helpers.lua:264: ';' expected near '__s128' at line 194'
exit code: 256
stack traceback:
test/unit/helpers.lua:748: in function 'itp_parent'
test/unit/helpers.lua:784: in function <test/unit/helpers.lua:774>
```
Since we don't use these types, they can be ignored to avoid LuaJIT's C
parser choking on them.
[6.5]: https://github.com/torvalds/linux/commit/224d80c584d3016cb8d83d1c33914fdd3508aa8c
59d5f692f removed cmake.deps/cmake/DownloadAndExtractFile.cmake and
support for USE_EXISTING_SRC_DIR. The Ubuntu nightly PPA still relies
on USE_EXISTING_SRC_DIR functionality since it can't access the network
during the build.
Supplying an empty value for ExternalProject_Add()'s URL value appears
to provide the needed mechanism to avoid re-downloading when the sources
are already present. This is undocumented behavior, though, so it may
break in the future.
Now, if USE_EXISTING_SRC_DIR is set, the ExternalProject's URL variable
is unset, preventing the download and erroring out if the source doesn't
actually exist.
Since we're not running tests or other things that are more sensitive to
changes in the VM environment, use ubuntu-latest to avoid the busy work
of updating the VM image.
The earlier vsnprintf() call checks whether str->items is NULL, sets of
the "possible NULL" spidey sense. kv_ensure_space() guarantees
str->items is non-NULL but since it doesn't use NULL checks to decide
whether to alloc, static analyzers can't tell this code path is safe.
register_cfunc allocates a ufunc_T, but doesn't store the pointer
anywhere before returning. The uf_name member variable is stored in a
hashtable and used to lookup the ufunc_T later, but that's too much for
Coverity to track.
Adding the annotation ensures that any new callers to register_cfunc
don't pop up as new "leaks" in the Coverity scans.
Coverity was reporting false positives, particularly around for non-NUL
terminated strings around uses of xmemdupz(). The updated model
ensures Coverity understands xmemdupz allocates an extra byte and sets it
to NUL as well as the main details of our other allocation related
wrappers.
James McCoy