math: fix frexp (don't check iptr when it's not specified)

[libc-test] / README

diff --git a/README b/README
index 856e2a8..02c8d2f 100644 (file)
--- a/README
+++ b/README
@@ -1,38 +1,22 @@
-simple libc tests based on the libc-testsuit
-and libc-bench of dalias
-see http://git.etalabs.net/cgi-bin/gitweb.cgi
+simple libc tests
+based on the libc-testsuit of dalias
+see http://git.musl-libc.org/cgit
 
 build tests:
-       cp Makefile.conf.def Makefile.conf
-       # edit Makefile.conf
+       cp dist/config.mak .
+       # edit config.mak
        make
 run tests:
-       ./t
-run benchmarks:
-       ./b
+       make run
 
-about the framework:
+framework:
 
-the only hook in the test framework is error(...) which
-prints a formatted message and sets the test to failed
-see common/test.h
+the convention is that each test is in a separate file
+at a path like src/directory/file.c with its own main
 
-benchmark functions are repeatedly called with increasing
-N until runing time is long enough, then time/N is printed
+the test should return 0 on success and non-0 on failure
 
-in the root dir make builds an executable with all tests
-in a src/* dir make builds only local tests
-see Makefile.inc
+to simplify things there is a Makefile.inc that can
+be included from src/directory/Makefile when the
+above convention is used in src/directory
 
-a test function looks like
-       void test_foo() {
-               if (foo != 42)
-                       error("foo=%d expected 42\n", foo);
-       }
-extern functions with name ~ /^test/ are recognized to be
-test functions
-see Makefile.inc
-
-edit the generated main.h to exclude certain tests
-see common/t.c for tests
-see common/b.c for benchmarks