posix: implement _POSIX_THREAD_SAFE_FUNCTIONS's time functions

[ycsin]

Add:

• asctime
• asctime_r
• localtime
• localtime_r
• ctime
• ctime_r

[cfriedt]

For the header file, this would probably work well.

Suggested change

	char *asctime(const struct tm *timeptr);
	struct tm *localtime(const time_t *timer);
	struct tm *localtime_r(const time_t *ZRESTRICT timer, struct tm *ZRESTRICT result);
	char *ctime(const time_t *clock);
	#if defined(CONFIG_COMMON_LIBC_ASCTIME_R) || defined(__DOXYGEN__)
	char *asctime_r(const struct tm *ZRESTRICT tp, char *ZRESTRICT buf);
	#endif /* CONFIG_COMMON_LIBC_ASCTIME_R */
	#if defined(CONFIG_COMMON_LIBC_CTIME_R) || defined(__DOXYGEN__)
	char *ctime_r(const time_t *clock, char *buf);
	#endif /* CONFIG_COMMON_LIBC_CTIME_R */
	char *asctime(const struct tm *timeptr);
	struct tm *localtime(const time_t *timer);
	char *ctime(const time_t *clock);
	#if defined(_POSIX_THREAD_SAFE_FUNCTIONS) || defined(__DOXYGEN__)
	char *asctime_r(const struct tm *ZRESTRICT tp, char *ZRESTRICT buf);
	char *ctime_r(const time_t *clock, char *buf);
	struct tm *localtime_r(const time_t *ZRESTRICT timer, struct tm *ZRESTRICT result);
	#endif

[ycsin]

How do I inject _POSIX_THREAD_SAFE_FUNCTIONS into this C header? I had a look at one of the picolibc time.h and they are not guarded

[cfriedt]

Darn - I just realized _POSIX_THREAD_SAFE_FUNCTIONS is an Option, and not an Option Group :(

Based on posixoptions(7), it seems to be a combination of POSIX_C_LANG_SUPPORT_R and POSIX_FILE_SYSTEM_R Option Groups

For reference
https://pubs.opengroup.org/onlinepubs/9799919799/xrat/V4_subprofiles.html

So likely we will need a new Kconfig.c_lang_r like this, approximately.

config POSIX_C_LANG_SUPPORT_R
    select COMMON_LIBC_ASCTIME_R
    select COMMON_LIBC_CTIME_R
    select COMMON_LIBC_GMTIME_R
    select COMMON_LIBC_LOCALTIME_R

With that, I think there are some nuances with how POSIX_THREAD_SAFE_FUNCTIONS should behave. Let me know if this makes sense.

Suggested change

	config POSIX_THREAD_SAFE_FUNCTIONS
	bool "POSIX thread-safe functions"
	select COMMON_LIBC_ASCTIME_R
	select COMMON_LIBC_GMTIME_R
	select COMMON_LIBC_LOCALTIME_R_UTC
	select COMMON_LIBC_CTIME_R
	help
	Select 'y' here to enable POSIX thread-safe functions including asctime_r(), ctime_r(),
	flockfile(), ftrylockfile(), funlockfile(), getc_unlocked(), getchar_unlocked(),
	config POSIX_THREAD_SAFE_FUNCTIONS
	bool "POSIX thread-safe functions"
	select POSIX_FILE_SYSTEM_R if POSIX_FILE_SYSTEM
	select POSIX_C_LANG_SUPPORT_R
	help
	Select 'y' here to enable POSIX thread-safe functions including asctime_r(), ctime_r(),
	flockfile(), ftrylockfile(), funlockfile(), getc_unlocked(), getchar_unlocked(),

[cfriedt]

I think this one can be abbreviated without the UTC.

From here

Unlike localtime(), the localtime_r() function is not required to set tzname. If localtime_r() sets tzname, it shall also set daylight and timezone. If localtime_r() does not set tzname, it shall not set daylight and shall not set timezone. If the tm structure member tm_zone is accessed after the value of TZ is subsequently modified, the behaviour is undefined.

So if our implementation of localtime_r() never sets the timezone (i.e. is UTC-only), it's still compliant. No special treatment necessary. My assumption is that most would probably like to avoid bringing timezone details into Zephyr.

Suggested change

	config COMMON_LIBC_LOCALTIME_R_UTC
	config COMMON_LIBC_LOCALTIME_R

[ycsin]

COMMON_LIBC_LOCALTIME_R_UTC is shared by both the localtime() & localtime_r(), I guess that's why I added that suffix to make it obvious that both functions only work for UTC, should I still remove the _UTC suffix? Or maybe I should create another Kconfig for localtime()?

[cfriedt]

Suggested change

	select COMMON_LIBC_LOCALTIME_R_UTC
	select COMMON_LIBC_LOCALTIME_R

[cfriedt]

Suggested change

	struct tm *localtime_r(const time_t *timer, struct tm *result)
	{
	return gmtime_r(timer, result);
	}
	#ifdef CONFIG_LIBC_TIME_LOCALTIME_R
	struct tm *localtime_r(const time_t *timer, struct tm *result)
	{
	return gmtime_r(timer, result);
	}
	#endif

[ycsin]

localtime_r() isn't a C function?

[cfriedt]

Pretty close - get some sleep and we'll check in again in the morning, your time.

[cfriedt]

@ycsin - I'm sorry that this PR (as well as several others you have made) have been delayed. I think this is partly due to false positives in Zephyr's CI.

As a workaround, feel free to include this commit in your PR.

[nashif]

DNM: has commit from PR that needs to be merged first.