Bernhard Reutner-Fischer
d61066fcfe
test: add API test
|
14 年 前 | |
|---|---|---|
| .. | ||
| API | 14 年 前 | |
| args | 15 年 前 | |
| assert | 15 年 前 | |
| build | 15 年 前 | |
| crypt | 15 年 前 | |
| ctype | 15 年 前 | |
| dlopen | 15 年 前 | |
| inet | 14 年 前 | |
| librt | 15 年 前 | |
| locale | 15 年 前 | |
| locale-mbwc | 15 年 前 | |
| malloc | 15 年 前 | |
| math | 14 年 前 | |
| misc | 15 年 前 | |
| mmap | 15 年 前 | |
| plt | 15 年 前 | |
| pthread | 15 年 前 | |
| pwd_grp | 15 年 前 | |
| regex | 15 年 前 | |
| rpc | 15 年 前 | |
| setjmp | 15 年 前 | |
| signal | 15 年 前 | |
| silly | 15 年 前 | |
| stat | 15 年 前 | |
| stdio | 15 年 前 | |
| stdlib | 15 年 前 | |
| string | 15 年 前 | |
| termios | 15 年 前 | |
| time | 15 年 前 | |
| unistd | 14 年 前 | |
| .gitignore | 15 年 前 | |
| Makefile | 15 年 前 | |
| README | 15 年 前 | |
| Rules.mak | 14 年 前 | |
| Test.mak | 15 年 前 | |
| test-skeleton.c | 15 年 前 | |
| testsuite.h | 18 年 前 | |
README
-----------
For: User
-----------
Following make targets are avaialable
make compile
This will compile and link the tests
make run
This will check for binaries, if they are not there it
will call 'compile' target, then it will execute all the tests.
make check
make all
This will build and run tests.
The following make variables may help you in testing:
- UCLIBC_ONLY - only run tests against uClibc
- GLIBC_ONLY - only run tests against glibc
- V / VERBOSE - run tests with a lot of output
- TEST_INSTALLED_UCLIBC - Test installed libraries
under /lib and /usr/lib.
- TIMEOUTFACTOR=nn - increase test timeout nn times.
At least REGEX_OLD + regex/tst-regex2 needs it increased.
So, to just run the uClibc tests, try this:
make check UCLIBC_ONLY=1
If you need to test just a subset of all test, delete subdirectories
you do not need.
As of 2009-07, build machinery does not track dependencies on uclibc.
If you edit a header and re-run "make compile", it does not re-install it
into ../install_dir. If you delete ../install_dir, "make compile"
rebuilds uclibc as needed and re-installs ../install_dir,
but still does not rebuild testcases.
(You can work around it by "touch */*.c" for now).
----------------
For: Developer
----------------
The structure of this test system is:
test/ toplevel dir containing common test code
test/Rules.mak Common build code
test/Test.mak Runtime test make code
test/subdir/ code specific to a subsystem is stored in a subdir
test/subdir/Makefile.in describe the tests to run
test/subdir/Makefile test entry point, includes needed upper-level
makefiles plus Makefile.in
test/subdir/*.c the tests
Each subdir has a Makefile (same for any subdir) that must include in strict order:
- the upper-level Rules.mak file
- the Makefile.in
- the upper-level Test.mak file
Makefile.in may be used to define the TESTS and TESTS_DISABLED variables.
If you do not, TESTS is built automatically based upon all the .c files in the subdir.
TESTS := foo
TESTS_DISABLED := bar
Each test must use a similar .c name; so the "foo" test needs a "foo.c".
Additionally, the following options further control specific test behavior:
CFLAGS_foo := extra cflags to use to compile test
DODIFF_foo := compare the output of the glibc and uClibc tests (see below)
LDFLAGS_foo := extra ldflags to use to link test
OPTS_foo := extra options to pass to test
RET_foo := expected exit code of test; default is 0
WRAPPER_foo := execute stuff just before test
Or to control all tests in a subdir:
EXTRA_CLEAN := extra files to remove in the clean target
EXTRA_DIRS := extra directories to remove in the clean target
EXTRA_CFLAGS := -DFOO
EXTRA_LDFLAGS := -lpthread
OPTS :=
WRAPPER :=
If you want to compare the output of a test with known good output, then just
create a local file named "foo.out.good" and the output generated by the test
"foo" will be automatically stored in "foo.out" and compared to "foo.out.good".
For: User
-----------
Following make targets are avaialable
make compile
This will compile and link the tests
make run
This will check for binaries, if they are not there it
will call 'compile' target, then it will execute all the tests.
make check
make all
This will build and run tests.
The following make variables may help you in testing:
- UCLIBC_ONLY - only run tests against uClibc
- GLIBC_ONLY - only run tests against glibc
- V / VERBOSE - run tests with a lot of output
- TEST_INSTALLED_UCLIBC - Test installed libraries
under /lib and /usr/lib.
- TIMEOUTFACTOR=nn - increase test timeout nn times.
At least REGEX_OLD + regex/tst-regex2 needs it increased.
So, to just run the uClibc tests, try this:
make check UCLIBC_ONLY=1
If you need to test just a subset of all test, delete subdirectories
you do not need.
As of 2009-07, build machinery does not track dependencies on uclibc.
If you edit a header and re-run "make compile", it does not re-install it
into ../install_dir. If you delete ../install_dir, "make compile"
rebuilds uclibc as needed and re-installs ../install_dir,
but still does not rebuild testcases.
(You can work around it by "touch */*.c" for now).
----------------
For: Developer
----------------
The structure of this test system is:
test/ toplevel dir containing common test code
test/Rules.mak Common build code
test/Test.mak Runtime test make code
test/subdir/ code specific to a subsystem is stored in a subdir
test/subdir/Makefile.in describe the tests to run
test/subdir/Makefile test entry point, includes needed upper-level
makefiles plus Makefile.in
test/subdir/*.c the tests
Each subdir has a Makefile (same for any subdir) that must include in strict order:
- the upper-level Rules.mak file
- the Makefile.in
- the upper-level Test.mak file
Makefile.in may be used to define the TESTS and TESTS_DISABLED variables.
If you do not, TESTS is built automatically based upon all the .c files in the subdir.
TESTS := foo
TESTS_DISABLED := bar
Each test must use a similar .c name; so the "foo" test needs a "foo.c".
Additionally, the following options further control specific test behavior:
CFLAGS_foo := extra cflags to use to compile test
DODIFF_foo := compare the output of the glibc and uClibc tests (see below)
LDFLAGS_foo := extra ldflags to use to link test
OPTS_foo := extra options to pass to test
RET_foo := expected exit code of test; default is 0
WRAPPER_foo := execute stuff just before test
Or to control all tests in a subdir:
EXTRA_CLEAN := extra files to remove in the clean target
EXTRA_DIRS := extra directories to remove in the clean target
EXTRA_CFLAGS := -DFOO
EXTRA_LDFLAGS := -lpthread
OPTS :=
WRAPPER :=
If you want to compare the output of a test with known good output, then just
create a local file named "foo.out.good" and the output generated by the test
"foo" will be automatically stored in "foo.out" and compared to "foo.out.good".