Use musl implementation of strftime rather than custom JS code. NFC (#21379)

This leads to some code size savings in several tests, reduces the
differences between standalone mode and non-standalone mode and reduces
the amount of custom JS code we need to maitain.

In simple microbenchmark that does nothing but export `strftime` we do
code size win at `-O0` but a regression in codesize at `-O2`:

```
$ wc -c old_O0.*
74026 old.js
12124 old.wasm
86150 total
$ wc -c new_O0.*
62120 new.js
22852 new.wasm
84972 total
```

```
$ wc -c old_O2.*
8492 old.js
 194 old.wasm
8686 total
$ wc -c new_O2.*
 5040 new.js
14291 new.wasm
19331 total
```

This regression comes from the fact that `snprintf` is used in the
implementation of `strftime` and its hard to optimize away the cost of
that function.

In practice I suspect that any application large enough to call
`strftime` is likely already directly or indirectly pulling in
`snprintf` so I would hope that this will be codesize win in practice
since it removes about 3k of JS.

Author: sbc100

ChangeLog.md

@@ -27,6 +27,8 @@ See docs/process.md for more on how version tagging works.
   `readAsync`. (#22080)
 - reference-types feature is now enabled by default in Emscripten, due to the
   upstream LLVM change (https://github.com/llvm/llvm-project/pull/93261).
+- Emscripten now uses `strftime` from musl rather than using a custom
+  JavaScript implementation. (#21379)
 
 3.1.61 - 05/31/24
 -----------------

src/library.js

@@ -728,266 +728,6 @@ addToLibrary({
     return newDate;
   },
 
-  // Note: this is not used in STANDALONE_WASM mode, because it is more
-  //       compact to do it in JS.
-  strftime__deps: ['$isLeapYear', '$arraySum', '$addDays', '$MONTH_DAYS_REGULAR', '$MONTH_DAYS_LEAP',
-                   '$intArrayFromString', '$writeArrayToMemory'
-  ],
-  strftime: (s, maxsize, format, tm) => {
-    // size_t strftime(char *restrict s, size_t maxsize, const char *restrict format, const struct tm *restrict timeptr);
-    // http://pubs.opengroup.org/onlinepubs/009695399/functions/strftime.html
-
-    var tm_zone = {{{ makeGetValue('tm', C_STRUCTS.tm.tm_zone, '*') }}};
-
-    var date = {
-      tm_sec: {{{ makeGetValue('tm', C_STRUCTS.tm.tm_sec, 'i32') }}},
-      tm_min: {{{ makeGetValue('tm', C_STRUCTS.tm.tm_min, 'i32') }}},
-      tm_hour: {{{ makeGetValue('tm', C_STRUCTS.tm.tm_hour, 'i32') }}},
-      tm_mday: {{{ makeGetValue('tm', C_STRUCTS.tm.tm_mday, 'i32') }}},
-      tm_mon: {{{ makeGetValue('tm', C_STRUCTS.tm.tm_mon, 'i32') }}},
-      tm_year: {{{ makeGetValue('tm', C_STRUCTS.tm.tm_year, 'i32') }}},
-      tm_wday: {{{ makeGetValue('tm', C_STRUCTS.tm.tm_wday, 'i32') }}},
-      tm_yday: {{{ makeGetValue('tm', C_STRUCTS.tm.tm_yday, 'i32') }}},
-      tm_isdst: {{{ makeGetValue('tm', C_STRUCTS.tm.tm_isdst, 'i32') }}},
-      tm_gmtoff: {{{ makeGetValue('tm', C_STRUCTS.tm.tm_gmtoff, LONG_TYPE) }}},
-      tm_zone: tm_zone ? UTF8ToString(tm_zone) : ''
-    };
-    {{{ from64('date.tm_gmtoff') }}}
-
-    var pattern = UTF8ToString(format);
-
-    // expand format
-    var EXPANSION_RULES_1 = {
-      '%D': '%m/%d/%y',                 // Equivalent to %m / %d / %y
-      '%F': '%Y-%m-%d',                 // Equivalent to %Y - %m - %d
-      '%h': '%b',                       // Equivalent to %b
-      '%r': '%I:%M:%S %p',              // Replaced by the time in a.m. and p.m. notation
-      '%R': '%H:%M',                    // Replaced by the time in 24-hour notation
-      '%T': '%H:%M:%S',                 // Replaced by the time
-      '%x': '%m/%d/%y',                 // Replaced by the locale's appropriate date representation
-      '%X': '%H:%M:%S',                 // Replaced by the locale's appropriate time representation
-      // Modified Conversion Specifiers
-      '%Ec': '%c',                      // Replaced by the locale's alternative appropriate date and time representation.
-      '%EC': '%C',                      // Replaced by the name of the base year (period) in the locale's alternative representation.
-      '%Ex': '%m/%d/%y',                // Replaced by the locale's alternative date representation.
-      '%EX': '%H:%M:%S',                // Replaced by the locale's alternative time representation.
-      '%Ey': '%y',                      // Replaced by the offset from %EC (year only) in the locale's alternative representation.
-      '%EY': '%Y',                      // Replaced by the full alternative year representation.
-      '%Od': '%d',                      // Replaced by the day of the month, using the locale's alternative numeric symbols, filled as needed with leading zeros if there is any alternative symbol for zero; otherwise, with leading <space> characters.
-      '%Oe': '%e',                      // Replaced by the day of the month, using the locale's alternative numeric symbols, filled as needed with leading <space> characters.
-      '%OH': '%H',                      // Replaced by the hour (24-hour clock) using the locale's alternative numeric symbols.
-      '%OI': '%I',                      // Replaced by the hour (12-hour clock) using the locale's alternative numeric symbols.
-      '%Om': '%m',                      // Replaced by the month using the locale's alternative numeric symbols.
-      '%OM': '%M',                      // Replaced by the minutes using the locale's alternative numeric symbols.
-      '%OS': '%S',                      // Replaced by the seconds using the locale's alternative numeric symbols.
-      '%Ou': '%u',                      // Replaced by the weekday as a number in the locale's alternative representation (Monday=1).
-      '%OU': '%U',                      // Replaced by the week number of the year (Sunday as the first day of the week, rules corresponding to %U ) using the locale's alternative numeric symbols.
-      '%OV': '%V',                      // Replaced by the week number of the year (Monday as the first day of the week, rules corresponding to %V ) using the locale's alternative numeric symbols.
-      '%Ow': '%w',                      // Replaced by the number of the weekday (Sunday=0) using the locale's alternative numeric symbols.
-      '%OW': '%W',                      // Replaced by the week number of the year (Monday as the first day of the week) using the locale's alternative numeric symbols.
-      '%Oy': '%y',                      // Replaced by the year (offset from %C ) using the locale's alternative numeric symbols.
-      '%c': '%a %b %d %H:%M:%S %Y',     // Replaced by the locale's appropriate date and time representation - e.g., Mon Aug  3 14:02:01 2013
-    };
-    for (var rule in EXPANSION_RULES_1) {
-      pattern = pattern.replace(new RegExp(rule, 'g'), EXPANSION_RULES_1[rule]);
-    }
-
-    var WEEKDAYS = ['Sunday', 'Monday', 'Tuesday', 'Wednesday', 'Thursday', 'Friday', 'Saturday'];
-    var MONTHS = ['January', 'February', 'March', 'April', 'May', 'June', 'July', 'August', 'September', 'October', 'November', 'December'];
-
-    function leadingSomething(value, digits, character) {
-      var str = typeof value == 'number' ? value.toString() : (value || '');
-      while (str.length < digits) {
-        str = character[0]+str;
-      }
-      return str;
-    }
-
-    function leadingNulls(value, digits) {
-      return leadingSomething(value, digits, '0');
-    }
-
-    function compareByDay(date1, date2) {
-      function sgn(value) {
-        return value < 0 ? -1 : (value > 0 ? 1 : 0);
-      }
-
-      var compare;
-      if ((compare = sgn(date1.getFullYear()-date2.getFullYear())) === 0) {
-        if ((compare = sgn(date1.getMonth()-date2.getMonth())) === 0) {
-          compare = sgn(date1.getDate()-date2.getDate());
-        }
-      }
-      return compare;
-    }
-
-    function getFirstWeekStartDate(janFourth) {
-      switch (janFourth.getDay()) {
-        case 0: // Sunday
-          return new Date(janFourth.getFullYear()-1, 11, 29);
-        case 1: // Monday
-          return janFourth;
-        case 2: // Tuesday
-          return new Date(janFourth.getFullYear(), 0, 3);
-        case 3: // Wednesday
-          return new Date(janFourth.getFullYear(), 0, 2);
-        case 4: // Thursday
-          return new Date(janFourth.getFullYear(), 0, 1);
-        case 5: // Friday
-          return new Date(janFourth.getFullYear()-1, 11, 31);
-        case 6: // Saturday
-          return new Date(janFourth.getFullYear()-1, 11, 30);
-      }
-    }
-
-    function getWeekBasedYear(date) {
-      var thisDate = addDays(new Date(date.tm_year+1900, 0, 1), date.tm_yday);
-
-      var janFourthThisYear = new Date(thisDate.getFullYear(), 0, 4);
-      var janFourthNextYear = new Date(thisDate.getFullYear()+1, 0, 4);
-
-      var firstWeekStartThisYear = getFirstWeekStartDate(janFourthThisYear);
-      var firstWeekStartNextYear = getFirstWeekStartDate(janFourthNextYear);
-
-      if (compareByDay(firstWeekStartThisYear, thisDate) <= 0) {
-        // this date is after the start of the first week of this year
-        if (compareByDay(firstWeekStartNextYear, thisDate) <= 0) {
-          return thisDate.getFullYear()+1;
-        }
-        return thisDate.getFullYear();
-      }
-      return thisDate.getFullYear()-1;
-    }
-
-    var EXPANSION_RULES_2 = {
-      '%a': (date) => WEEKDAYS[date.tm_wday].substring(0,3) ,
-      '%A': (date) => WEEKDAYS[date.tm_wday],
-      '%b': (date) => MONTHS[date.tm_mon].substring(0,3),
-      '%B': (date) => MONTHS[date.tm_mon],
-      '%C': (date) => {
-        var year = date.tm_year+1900;
-        return leadingNulls((year/100)|0,2);
-      },
-      '%d': (date) => leadingNulls(date.tm_mday, 2),
-      '%e': (date) => leadingSomething(date.tm_mday, 2, ' '),
-      '%g': (date) => {
-        // %g, %G, and %V give values according to the ISO 8601:2000 standard week-based year.
-        // In this system, weeks begin on a Monday and week 1 of the year is the week that includes
-        // January 4th, which is also the week that includes the first Thursday of the year, and
-        // is also the first week that contains at least four days in the year.
-        // If the first Monday of January is the 2nd, 3rd, or 4th, the preceding days are part of
-        // the last week of the preceding year; thus, for Saturday 2nd January 1999,
-        // %G is replaced by 1998 and %V is replaced by 53. If December 29th, 30th,
-        // or 31st is a Monday, it and any following days are part of week 1 of the following year.
-        // Thus, for Tuesday 30th December 1997, %G is replaced by 1998 and %V is replaced by 01.
-
-        return getWeekBasedYear(date).toString().substring(2);
-      },
-      '%G': getWeekBasedYear,
-      '%H': (date) => leadingNulls(date.tm_hour, 2),
-      '%I': (date) => {
-        var twelveHour = date.tm_hour;
-        if (twelveHour == 0) twelveHour = 12;
-        else if (twelveHour > 12) twelveHour -= 12;
-        return leadingNulls(twelveHour, 2);
-      },
-      '%j': (date) => {
-        // Day of the year (001-366)
-        return leadingNulls(date.tm_mday + arraySum(isLeapYear(date.tm_year+1900) ? MONTH_DAYS_LEAP : MONTH_DAYS_REGULAR, date.tm_mon-1), 3);
-      },
-      '%m': (date) => leadingNulls(date.tm_mon+1, 2),
-      '%M': (date) => leadingNulls(date.tm_min, 2),
-      '%n': () => '\n',
-      '%p': (date) => {
-        if (date.tm_hour >= 0 && date.tm_hour < 12) {
-          return 'AM';
-        }
-        return 'PM';
-      },
-      '%S': (date) => leadingNulls(date.tm_sec, 2),
-      '%t': () => '\t',
-      '%u': (date) => date.tm_wday || 7,
-      '%U': (date) => {
-        var days = date.tm_yday + 7 - date.tm_wday;
-        return leadingNulls(Math.floor(days / 7), 2);
-      },
-      '%V': (date) => {
-        // Replaced by the week number of the year (Monday as the first day of the week)
-        // as a decimal number [01,53]. If the week containing 1 January has four
-        // or more days in the new year, then it is considered week 1.
-        // Otherwise, it is the last week of the previous year, and the next week is week 1.
-        // Both January 4th and the first Thursday of January are always in week 1. [ tm_year, tm_wday, tm_yday]
-        var val = Math.floor((date.tm_yday + 7 - (date.tm_wday + 6) % 7 ) / 7);
-        // If 1 Jan is just 1-3 days past Monday, the previous week
-        // is also in this year.
-        if ((date.tm_wday + 371 - date.tm_yday - 2) % 7 <= 2) {
-          val++;
-        }
-        if (!val) {
-          val = 52;
-          // If 31 December of prev year a Thursday, or Friday of a
-          // leap year, then the prev year has 53 weeks.
-          var dec31 = (date.tm_wday + 7 - date.tm_yday - 1) % 7;
-          if (dec31 == 4 || (dec31 == 5 && isLeapYear(date.tm_year%400-1))) {
-            val++;
-          }
-        } else if (val == 53) {
-          // If 1 January is not a Thursday, and not a Wednesday of a
-          // leap year, then this year has only 52 weeks.
-          var jan1 = (date.tm_wday + 371 - date.tm_yday) % 7;
-          if (jan1 != 4 && (jan1 != 3 || !isLeapYear(date.tm_year)))
-            val = 1;
-        }
-        return leadingNulls(val, 2);
-      },
-      '%w': (date) => date.tm_wday,
-      '%W': (date) => {
-        var days = date.tm_yday + 7 - ((date.tm_wday + 6) % 7);
-        return leadingNulls(Math.floor(days / 7), 2);
-      },
-      '%y': (date) => {
-        // Replaced by the last two digits of the year as a decimal number [00,99]. [ tm_year]
-        return (date.tm_year+1900).toString().substring(2);
-      },
-      // Replaced by the year as a decimal number (for example, 1997). [ tm_year]
-      '%Y': (date) => date.tm_year+1900,
-      '%z': (date) => {
-        // Replaced by the offset from UTC in the ISO 8601:2000 standard format ( +hhmm or -hhmm ).
-        // For example, "-0430" means 4 hours 30 minutes behind UTC (west of Greenwich).
-        var off = date.tm_gmtoff;
-        var ahead = off >= 0;
-        off = Math.abs(off) / 60;
-        // convert from minutes into hhmm format (which means 60 minutes = 100 units)
-        off = (off / 60)*100 + (off % 60);
-        return (ahead ? '+' : '-') + String("0000" + off).slice(-4);
-      },
-      '%Z': (date) => date.tm_zone,
-      '%%': () => '%'
-    };
-
-    // Replace %% with a pair of NULLs (which cannot occur in a C string), then
-    // re-inject them after processing.
-    pattern = pattern.replace(/%%/g, '\0\0')
-    for (var rule in EXPANSION_RULES_2) {
-      if (pattern.includes(rule)) {
-        pattern = pattern.replace(new RegExp(rule, 'g'), EXPANSION_RULES_2[rule](date));
-      }
-    }
-    pattern = pattern.replace(/\0\0/g, '%')
-
-    var bytes = intArrayFromString(pattern, false);
-    if (bytes.length > maxsize) {
-      return 0;
-    }
-
-    writeArrayToMemory(bytes, s);
-    return bytes.length-1;
-  },
-  strftime_l__deps: ['strftime'],
-  strftime_l: (s, maxsize, format, tm, loc) => {
-    return _strftime(s, maxsize, format, tm); // no locale support yet
-  },
-
   strptime__deps: ['$isLeapYear', '$arraySum', '$addDays', '$MONTH_DAYS_REGULAR', '$MONTH_DAYS_LEAP',
                    '$jstoi_q', '$intArrayFromString' ],
   strptime: (buf, format, tm) => {

src/library_sigs.js

@@ -1536,8 +1536,6 @@ sigs = {
   rotozoomSurface__sig: 'ppddi',
   sched_yield__sig: 'i',
   setprotoent__sig: 'vi',
-  strftime__sig: 'ppppp',
-  strftime_l__sig: 'pppppp',
   strptime__sig: 'pppp',
   strptime_l__sig: 'ppppp',
   uuid_clear__sig: 'vp',

system/lib/libc/emscripten_time.c

@@ -15,12 +15,6 @@
 
 #include "emscripten_internal.h"
 
-// Replaces musl's __tz.c
-
-weak long  timezone = 0;
-weak int   daylight = 0;
-weak char *tzname[2] = { 0, 0 };
-
 weak clock_t __clock() {
   static thread_local double start = 0;
   if (!start) {

system/lib/libc/musl/src/time/__tz.c

@@ -9,6 +9,12 @@
 #include "lock.h"
 #include "fork_impl.h"
 
+#ifdef __EMSCRIPTEN__
+#include <stdbool.h>
+#include <pthread.h>
+#include "emscripten_internal.h"
+#endif
+
 #define malloc __libc_malloc
 #define calloc undef
 #define realloc undef
@@ -39,6 +45,7 @@ static size_t old_tz_size = sizeof old_tz_buf;
 static volatile int lock[1];
 volatile int *const __timezone_lockptr = lock;
 
+#ifndef __EMSCRIPTEN__
 static int getint(const char **p)
 {
 	unsigned x;
@@ -122,9 +129,24 @@ static size_t zi_dotprod(const unsigned char *z, const unsigned char *v, size_t
 	}
 	return y;
 }
+#endif
 
 static void do_tzset()
 {
+#ifdef __EMSCRIPTEN__
+	static pthread_mutex_t lock = PTHREAD_MUTEX_INITIALIZER;
+	static _Atomic bool done_init = false;
+	if (!done_init) {
+		pthread_mutex_lock(&lock);
+		if (!done_init) {
+			_tzset_js(&timezone, &daylight, std_name, dst_name);
+			__tzname[0] = std_name;
+			__tzname[1] = dst_name;
+			done_init = true;
+		}
+		pthread_mutex_unlock(&lock);
+	}
+#else
 	char buf[NAME_MAX+25], *pathname=buf+24;
 	const char *try, *s, *p;
 	const unsigned char *map = 0;
@@ -255,8 +277,10 @@ static void do_tzset()
 
 	if (*s == ',') s++, getrule(&s, r0);
 	if (*s == ',') s++, getrule(&s, r1);
+#endif
 }
 
+#ifndef __EMSCRIPTEN__
 /* Search zoneinfo rules to find the one that applies to the given time,
  * and determine alternate opposite-DST-status rule that may be needed. */
 
@@ -416,6 +440,7 @@ void __secs_to_zone(long long t, int local, int *isdst, long *offset, long *oppo
 	*zonename = __tzname[1];
 	UNLOCK(lock);
 }
+#endif
 
 static void __tzset()
 {