| vtorri | |
| raster | |
| q66 | |
| stefan_schmidt |
In some platforms such as Windows, there are cases when command line
length is too short even to run Eolian tests (depending on which
directory EFL repository is cloned into). This commit makes enables the
usage of response files, which contains each of the args that would go
into command line splitted one per line.
Ref T8768
meson test -C build --print-errorlogs
| No Linters Available |
| No Unit Test Coverage |
| Buildable 17326 | |
| Build 11589: arc lint + arc unit |
| src/bin/eolian/main.c | ||
|---|---|---|
| 497 | we don't use camel case for the name of types, we add _ to separate parts. Like : Cmd_Line_Options | |
| src/lib/eina/eina_file_win32.c | ||
| 109 ↗ | (On Diff #31022) | maybe another patch for this one ? The code is not wrong as INVALID_FILE_ATTRIBUTES is equal to 0xffffffff, but you're right, better using this name. |
Fix CmdLineOptions case to Cmd_Line_Options and undo
INVALID_FILE_ATTRIBUTES change (which will be splitted into another patch).
Applies now, but shows errors when building:
[1/3252] Compiling C object 'src/bin/eolian/70b716a@@eolian_gen@exe/main.c.o'
../src/bin/eolian/main.c: In function ‘_read_options_from_response_file’:
../src/bin/eolian/main.c:524:22: warning: comparison of integer expressions of different signedness: ‘int’ and ‘size_t’ {aka ‘long unsigned int’} [-Wsign-compare]
524 | for (int i = 0; i < file_length; ++i)
| ^
../src/bin/eolian/main.c:543:22: warning: comparison of integer expressions of different signedness: ‘int’ and ‘size_t’ {aka ‘long unsigned int’} [-Wsign-compare]
543 | for (int i = 0; i < file_length; ++i)
| ^
../src/bin/eolian/main.c:548:23: warning: comparison of integer expressions of different signedness: ‘int’ and ‘size_t’ {aka ‘long unsigned int’} [-Wsign-compare]
548 | while (i < file_length && contents[i] == '\0')
| ^
../src/bin/eolian/main.c:553:20: warning: comparison of integer expressions of different signedness: ‘int’ and ‘size_t’ {aka ‘long unsigned int’} [-Wsign-compare]
553 | if (i >= file_length)
| ^~Just my two cents: I think the tests should unconditionally run with the respones files, and not only if PATH_MAX is too small, this way we can ensure that the respone file is tested on any maschine that runs the tests.
Its even more than that. When running the test suite I get the following:
../src/tests/eolian/eolian_generation.c: In function ‘_eolian_gen_execute’:
../src/tests/eolian/eolian_generation.c:127:67: error: ‘type’ undeclared (first use in this function)
127 | int failed = _create_response_file(rsp_path, eo_filename, type, output_filename);
| ^~~~
../src/tests/eolian/eolian_generation.c:127:67: note: each undeclared identifier is reported only once for each function it appears in
<command-line>: warning: ‘%s’ directive output may be truncated writing up to 4095 bytes into a region of size 4034 [-Wformat-truncation=]
<command-line>: note: in definition of macro ‘EOLIAN_GEN’
../src/tests/eolian/eolian_generation.c:123:9: note: ‘snprintf’ output between 64 and 4159 bytes into a destination of size 4096
123 | snprintf(command, PATH_MAX,
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~
124 | EOLIAN_GEN" -r \"%s\"",
| ~~~~~~~~~~~~~~~~~~~~~~~
125 | rsp_path);*Facepalm* My bad, totally didn't notice that. Gonna fix that.
Oh, right. Well, gonna test a little better before updating.
Fix wrong cherry-pick resolution (i.e. use options again instead of type, which was from another commit that is not on upstream).
You mean adding a test for that or actually changing all tests to use a response file?
While the tests build error is gone the different signedness warnings are still there