[casaroli]

depends-on: apache/nuttx/pull/21595

Part of apache/nuttx#21467.

Summary

Tests of printf() for the merged apache/nuttx#21505 and apache/nuttx#21507, and for apache/nuttx#21595. Four commits:

1. printftest follows the printf options. It expected the output of the Default profile, so it failed with the other profiles of libc and when one printf option was turned off. Each case now follows LIBC_FLOATINGPOINT, LIBC_NUMBERED_ARGS, LIBC_PRINTF_ALT_FORM, LIBC_PRINTF_INT_PRECISION, LIBC_PRINTF_OCTAL (%o prints "?"), LIBC_PRINTF_LEFT_JUSTIFY (the - flag pads on the left), LIBC_PRINTF_HEXFLOAT, LIBC_PRINTF_C23 and LIBC_PRINTF_PERCENT_N. LIBC_PRINTF_SHORT and LIBC_PRINTF_LONG_LONG change no case: %hhn and %lln store at their size with both settings. New cases: the sign of -0.0, %a and %A, %b and %B, and the wN length. A precision of INT_MIN from * is no precision, so %.*d|%.*s| expects 1|ab| in all profiles.
2. printftest: cases for %#g, which keeps the trailing zeros (%#g of 1.0 is 1.00000, %#.0g of 1.0 is 1.), and for a negative precision from *, which is no precision (%.*s with -2 prints the whole string, %0*.*d keeps the 0 flag).
3. A new program, printfengine, checks the digits of %e, %f and %g: ties of values that are exact in binary, %.0f rounds, the form of %g, the sign, infinity and NaN, the limits of binary64 and 17 correct digits. With LIBC_PRINTF_EXACT it checks all digits of a double, and with LIBC_PRINTF_EXACT_LDOUBLE the range and the digits of a long double. It follows printftest: the same layout, and Make and CMake files.
4. printftest: %n prints nothing also when LIBC_PRINTF_PERCENT_N is off (libs/libc/stream: Print nothing for %n when LIBC_PRINTF_PERCENT_N is off. nuttx#21595), where it printed ? before. The test expects the same text with the option on or off, and checks the store only when the option is on. New case: snprintf(NULL, 0, "ab%n|%s", &n, "str") returns 6.

Impact

testing/libc/printftest/printftest_main.c, and the new program in testing/libc/printfengine with the option CONFIG_TESTING_PRINTFENGINE, off by default. It depends on LIBC_FLOATINGPOINT. No change to any other code.

Commit 4 needs apache/nuttx#21595: on master printftest fails at %n with LIBC_PRINTF_PERCENT_N off. So this PR must merge after it.

Testing

qemu-armv8a under QEMU with HVF, nuttx apache/nuttx#21595 (9ae1b2e6d85, on master d1191e0ad01), printftest, printfengine, then ostest; each run passes and ostest exits with status 0:

build	printf settings
qemu-armv8a:knsh (kernel)	Default
qemu-armv8a:knsh	Full
qemu-armv8a:knsh	Minimal
qemu-armv8a:knsh	Default, LIBC_PRINTF_OCTAL=n, LIBC_PRINTF_LEFT_JUSTIFY=n
qemu-armv8a:knsh	Full, LIBC_PRINTF_PERCENT_N=y
qemu-armv8a:knsh	Minimal, LIBC_PRINTF_PERCENT_N=y, LIBC_PRINTF_OCTAL=n, LIBC_PRINTF_LEFT_JUSTIFY=n
qemu-armv8a:nsh (flat)	Default
qemu-armv8a:nsh	Full, LIBC_PRINTF_PERCENT_N=y

printfengine needs floating point, so it does not run with Minimal.

On nuttx master d1191e0ad01 (qemu-armv8a:knsh, Default) printftest fails at commit 4, as it must: "ab%n|%s" returned 7: "ab?|str", expected "ab|str". printfengine passes there. With LIBC_PRINTF_PERCENT_N=y both pass on master.

Builds with -Werror and both programs on: qemu-armv8a:nsh and rv-virt:nsh64, each with Default and with Full and LIBC_PRINTF_PERCENT_N=y.

tools/checkpatch.sh -c -u -m -g passes.

Not run: a CMake build, a protected build, hardware.