Refactoring time-related default port implementations

[akosthekiss]

In the non-Windows code paths:

• New approach to compute TZA without the need for GNU-specific
  struct tm.tm_gmtoff.
• Always using usleep to sleep. (No real need for nanosleep as
  port API has sleep granularity of milliseconds.)
• Not checking for "time.h" at build configuration time as that
  header is mandated by the C standard.
• Not checking for "unistd.h" at build configuration time as that
  header is mandated by the POSIX standard (the default port is
  targeting POSIX systems -- and Windows).
• Fixing some macro guards.

JerryScript-DCO-1.0-Signed-off-by: Akos Kiss akiss@inf.u-szeged.hu

[akosthekiss]

I have to admit that I'm not sure whether existing test cases cover these routines properly. I've partially validated the TZA computation with the C code below:

#include <stdio.h>
#include <time.h>

double tza_old(time_t time_utc) {
  tzset(); // should be executed only once?

  // time in broken-down structure, expressed in local time zone
  struct tm tm_local;
  localtime_r(&time_utc, &tm_local);

  // seconds east of UTC (from GNU-specific structure field)
  return (double)tm_local.tm_gmtoff;
}

double tza_new(time_t time_utc) {
  // time in broken-down structure, expressed as UTC time
  struct tm tm_utc;
  gmtime_r(&time_utc, &tm_utc);

  // time converted back to seconds since Epoch UTC
  tm_utc.tm_isdst = -1; // if not overridden, DST will not be taken into account
  time_t time_local = mktime(&tm_utc);

  // seconds east of UTC (from simple time differences)
  return difftime(time_utc, time_local);
}

int main() {
  time_t now_utc = time(NULL); // seconds since Epoch UTC
  time_t past_utc = now_utc - 182L * 24 * 60 * 60; // cca. half a year ago

  printf("tza-old now: %lf\n", tza_old(now_utc));
  printf("tza-old past: %lf\n", tza_old(past_utc));

  printf("tza-new now: %lf\n", tza_new(now_utc));
  printf("tza-new past: %lf\n", tza_new(past_utc));

  return 0;
}

Built and ran on Ubuntu 18.04.5 LTS (Linux 4.15.0-130-generic x86_64) as:

$ gcc tza.c -o tza
$ ./tza 
tza-old now: 3600.000000
tza-old past: 7200.000000
tza-new now: 3600.000000
tza-new past: 7200.000000

[lygstate]

LGTM

[lygstate]

@akosthekiss Does this what you want?

lygstate@DESKTOP-94PU0GB:/mnt/c/work/study/languages/typescript/jerryscript$ ./a.out
tza-old now: -28800.000000
tza-old past: -25200.000000
tza-new now: -28800.000000
tza-new past: -25200.000000

[akosthekiss]

@lygstate Thanks. Depends on your actual time zone. Looks like GMT-8: PST? Validates part of the code. But the is_utc=false logic of the TZA port function is still hard to check whether it works OK. (The above C code does not validate that at all.)

[lygstate]

Yeap, it's GMT-8

[lygstate]

@lygstate Thanks. Depends on your actual time zone. Looks like GMT-8: PST? Validates part of the code. But the is_utc=false logic of the TZA port function is still hard to check whether it works OK. (The above C code does not validate that at all.)

If I read the code correctly, the is_utc option didn't considerated on _WIN32

[lygstate]

I have a question, If the OS provide what's the current time zone is, can we use the timezone info directly implement jerry_port_get_local_time_zone_adjustment ?

[rerobika]

LGTM after the style fix

[rerobika]

Please add the missing comments to preprocessor directives. (here and below)

[akosthekiss]

Added the pre-processor directive comments. PR updated. Thanks for the review.

[zherczeg]

LGTM