util-lib: make timestamp generation and parsing reversible (#3869)

This patch improves parsing and generation of timestamps and calendar
specifications in two ways:

- The week day is now always printed in the abbreviated English form, instead
  of the locale's setting. This makes sure we can always parse the week day
  again, even if the locale is changed. Given that we don't follow locale
  settings for printing timestamps in any other way either (for example, we
  always use 24h syntax in order to make uniform parsing possible), it only
  makes sense to also stick to a generic, non-localized form for the timestamp,
  too.

- When parsing a timestamp, the local timezone (in its DST or non-DST name)
  may be specified, in addition to "UTC". Other timezones are still not
  supported however (not because we wouldn't want to, but mostly because libc
  offers no nice API for that). In itself this brings no new features, however
  it ensures that any locally formatted timestamp's timezone is also parsable
  again.

These two changes ensure that the output of format_timestamp() may always be
passed to parse_timestamp() and results in the original input. The related
flavours for usec/UTC also work accordingly. Calendar specifications are
extended in a similar way.

The man page is updated accordingly, in particular this removes the claim that
timestamps systemd prints wouldn't be parsable by systemd. They are now.

The man page previously showed invalid timestamps as examples. This has been
removed, as the man page shouldn't be a unit test, where such negative examples
would be useful. The man page also no longer mentions the names of internal
functions, such as format_timestamp_us() or UNIX error codes such as EINVAL.

4 files changed, 168 insertions, 26 deletions

diff --git a/src/basic/calendarspec.c b/src/basic/calendarspec.c
index e4cfab364e..fda293fcb9 100644
--- a/src/basic/calendarspec.c
+++ b/src/basic/calendarspec.c
@@ -302,6 +302,17 @@ int calendar_spec_to_string(const CalendarSpec *c, char **p) {
 
         if (c->utc)
                 fputs(" UTC", f);
+        else if (IN_SET(c->dst, 0, 1)) {
+
+                /* If daylight saving is explicitly on or off, let's show the used timezone. */
+
+                tzset();
+
+                if (!isempty(tzname[c->dst])) {
+                        fputc(' ', f);
+                        fputs(tzname[c->dst], f);
+                }
+        }
 
         r = fflush_and_check(f);
         if (r < 0) {
@@ -747,9 +758,9 @@ fail:
 }
 
 int calendar_spec_from_string(const char *p, CalendarSpec **spec) {
+        const char *utc;
         CalendarSpec *c;
         int r;
-        const char *utc;
 
         assert(p);
         assert(spec);
@@ -760,11 +771,39 @@ int calendar_spec_from_string(const char *p, CalendarSpec **spec) {
         c = new0(CalendarSpec, 1);
         if (!c)
                 return -ENOMEM;
+        c->dst = -1;
 
         utc = endswith_no_case(p, " UTC");
         if (utc) {
                 c->utc = true;
                 p = strndupa(p, utc - p);
+        } else {
+                const char *e = NULL;
+                int j;
+
+                tzset();
+
+                /* Check if the local timezone was specified? */
+                for (j = 0; j <= 1; j++) {
+                        if (isempty(tzname[j]))
+                                continue;
+
+                        e = endswith_no_case(p, tzname[j]);
+                        if(!e)
+                                continue;
+                        if (e == p)
+                                continue;
+                        if (e[-1] != ' ')
+                                continue;
+
+                        break;
+                }
+
+                /* Found one of the two timezones specified? */
+                if (IN_SET(j, 0, 1)) {
+                        p = strndupa(p, e - p - 1);
+                        c->dst = j;
+                }
         }
 
         if (strcaseeq(p, "minutely")) {
@@ -1017,7 +1056,7 @@ static int find_next(const CalendarSpec *spec, struct tm *tm, usec_t *usec) {
         for (;;) {
                 /* Normalize the current date */
                 (void) mktime_or_timegm(&c, spec->utc);
-                c.tm_isdst = -1;
+                c.tm_isdst = spec->dst;
 
                 c.tm_year += 1900;
                 r = find_matching_component(spec->year, &c.tm_year);
diff --git a/src/basic/calendarspec.h b/src/basic/calendarspec.h
index f6472c1244..c6087228fd 100644
--- a/src/basic/calendarspec.h
+++ b/src/basic/calendarspec.h
@@ -37,6 +37,7 @@ typedef struct CalendarComponent {
 typedef struct CalendarSpec {
         int weekdays_bits;
         bool utc;
+        int dst;
 
         CalendarComponent *year;
         CalendarComponent *month;
diff --git a/src/basic/time-util.c b/src/basic/time-util.c
index 24e681bf85..0ef1f6393e 100644
--- a/src/basic/time-util.c
+++ b/src/basic/time-util.c
@@ -254,32 +254,95 @@ struct timeval *timeval_store(struct timeval *tv, usec_t u) {
         return tv;
 }
 
-static char *format_timestamp_internal(char *buf, size_t l, usec_t t,
-                                       bool utc, bool us) {
+static char *format_timestamp_internal(
+                char *buf,
+                size_t l,
+                usec_t t,
+                bool utc,
+                bool us) {
+
+        /* The weekdays in non-localized (English) form. We use this instead of the localized form, so that our
+         * generated timestamps may be parsed with parse_timestamp(), and always read the same. */
+        static const char * const weekdays[] = {
+                [0] = "Sun",
+                [1] = "Mon",
+                [2] = "Tue",
+                [3] = "Wed",
+                [4] = "Thu",
+                [5] = "Fri",
+                [6] = "Sat",
+        };
+
         struct tm tm;
         time_t sec;
-        int k;
+        size_t n;
 
         assert(buf);
-        assert(l > 0);
 
+        if (l <
+            3 +                  /* week day */
+            1 + 10 +             /* space and date */
+            1 + 8 +              /* space and time */
+            (us ? 1 + 6 : 0) +   /* "." and microsecond part */
+            1 + 1 +              /* space and shortest possible zone */
+            1)
+                return NULL; /* Not enough space even for the shortest form. */
         if (t <= 0 || t == USEC_INFINITY)
+                return NULL; /* Timestamp is unset */
+
+        sec = (time_t) (t / USEC_PER_SEC); /* Round down */
+        if ((usec_t) sec != (t / USEC_PER_SEC))
+                return NULL; /* overflow? */
+
+        if (!localtime_or_gmtime_r(&sec, &tm, utc))
                 return NULL;
 
-        sec = (time_t) (t / USEC_PER_SEC);
-        localtime_or_gmtime_r(&sec, &tm, utc);
+        /* Start with the week day */
+        assert((size_t) tm.tm_wday < ELEMENTSOF(weekdays));
+        memcpy(buf, weekdays[tm.tm_wday], 4);
 
-        if (us)
-                k = strftime(buf, l, "%a %Y-%m-%d %H:%M:%S", &tm);
-        else
-                k = strftime(buf, l, "%a %Y-%m-%d %H:%M:%S %Z", &tm);
+        /* Add the main components */
+        if (strftime(buf + 3, l - 3, " %Y-%m-%d %H:%M:%S", &tm) <= 0)
+                return NULL; /* Doesn't fit */
 
-        if (k <= 0)
-                return NULL;
+        /* Append the microseconds part, if that's requested */
         if (us) {
-                snprintf(buf + strlen(buf), l - strlen(buf), ".%06llu", (unsigned long long) (t % USEC_PER_SEC));
-                if (strftime(buf + strlen(buf), l - strlen(buf), " %Z", &tm) <= 0)
-                        return NULL;
+                n = strlen(buf);
+                if (n + 8 > l)
+                        return NULL; /* Microseconds part doesn't fit. */
+
+                sprintf(buf + n, ".%06llu", (unsigned long long) (t % USEC_PER_SEC));
+        }
+
+        /* Append the timezone */
+        n = strlen(buf);
+        if (utc) {
+                /* If this is UTC then let's explicitly use the "UTC" string here, because gmtime_r() normally uses the
+                 * obsolete "GMT" instead. */
+                if (n + 5 > l)
+                        return NULL; /* "UTC" doesn't fit. */
+
+                strcpy(buf + n, " UTC");
+
+        } else if (!isempty(tm.tm_zone)) {
+                size_t tn;
+
+                /* An explicit timezone is specified, let's use it, if it fits */
+                tn = strlen(tm.tm_zone);
+                if (n + 1 + tn + 1 > l) {
+                        /* The full time zone does not fit in. Yuck. */
+
+                        if (n + 1 + _POSIX_TZNAME_MAX + 1 > l)
+                                return NULL; /* Not even enough space for the POSIX minimum (of 6)? In that case, complain that it doesn't fit */
+
+                        /* So the time zone doesn't fit in fully, but the caller passed enough space for the POSIX
+                         * minimum time zone length. In this case suppress the timezone entirely, in order not to dump
+                         * an overly long, hard to read string on the user. This should be safe, because the user will
+                         * assume the local timezone anyway if none is shown. And so does parse_timestamp(). */
+                } else {
+                        buf[n++] = ' ';
+                        strcpy(buf + n, tm.tm_zone);
+                }
         }
 
         return buf;
@@ -539,12 +602,11 @@ int parse_timestamp(const char *t, usec_t *usec) {
                 { "Sat",       6 },
         };
 
-        const char *k;
-        const char *utc;
+        const char *k, *utc, *tzn = NULL;
         struct tm tm, copy;
         time_t x;
         usec_t x_usec, plus = 0, minus = 0, ret;
-        int r, weekday = -1;
+        int r, weekday = -1, dst = -1;
         unsigned i;
 
         /*
@@ -609,15 +671,55 @@ int parse_timestamp(const char *t, usec_t *usec) {
                 goto finish;
         }
 
+        /* See if the timestamp is suffixed with UTC */
         utc = endswith_no_case(t, " UTC");
         if (utc)
                 t = strndupa(t, utc - t);
+        else {
+                const char *e = NULL;
+                int j;
+
+                tzset();
+
+                /* See if the timestamp is suffixed by either the DST or non-DST local timezone. Note that we only
+                 * support the local timezones here, nothing else. Not because we wouldn't want to, but simply because
+                 * there are no nice APIs available to cover this. By accepting the local time zone strings, we make
+                 * sure that all timestamps written by format_timestamp() can be parsed correctly, even though we don't
+                 * support arbitrary timezone specifications.  */
+
+                for (j = 0; j <= 1; j++) {
+
+                        if (isempty(tzname[j]))
+                                continue;
+
+                        e = endswith_no_case(t, tzname[j]);
+                        if (!e)
+                                continue;
+                        if (e == t)
+                                continue;
+                        if (e[-1] != ' ')
+                                continue;
+
+                        break;
+                }
 
-        x = ret / USEC_PER_SEC;
+                if (IN_SET(j, 0, 1)) {
+                        /* Found one of the two timezones specified. */
+                        t = strndupa(t, e - t - 1);
+                        dst = j;
+                        tzn = tzname[j];
+                }
+        }
+
+        x = (time_t) (ret / USEC_PER_SEC);
         x_usec = 0;
 
-        assert_se(localtime_or_gmtime_r(&x, &tm, utc));
-        tm.tm_isdst = -1;
+        if (!localtime_or_gmtime_r(&x, &tm, utc))
+                return -EINVAL;
+
+        tm.tm_isdst = dst;
+        if (tzn)
+                tm.tm_zone = tzn;
 
         if (streq(t, "today")) {
                 tm.tm_sec = tm.tm_min = tm.tm_hour = 0;
@@ -634,7 +736,6 @@ int parse_timestamp(const char *t, usec_t *usec) {
                 goto from_tm;
         }
 
-
         for (i = 0; i < ELEMENTSOF(day_nr); i++) {
                 size_t skip;
 
@@ -727,7 +828,6 @@ parse_usec:
                         return -EINVAL;
 
                 x_usec = add;
-
         }
 
 from_tm:
diff --git a/src/basic/time-util.h b/src/basic/time-util.h
index 1b058f0e49..99be5ce6ee 100644
--- a/src/basic/time-util.h
+++ b/src/basic/time-util.h
@@ -68,7 +68,9 @@ typedef struct triple_timestamp {
 #define USEC_PER_YEAR ((usec_t) (31557600ULL*USEC_PER_SEC))
 #define NSEC_PER_YEAR ((nsec_t) (31557600ULL*NSEC_PER_SEC))
 
-#define FORMAT_TIMESTAMP_MAX ((4*4+1)+11+9+4+1) /* weekdays can be unicode */
+/* We assume a maximum timezone length of 6. TZNAME_MAX is not defined on Linux, but glibc internally initializes this
+ * to 6. Let's rely on that. */
+#define FORMAT_TIMESTAMP_MAX (3+1+10+1+8+1+6+1+6+1)
 #define FORMAT_TIMESTAMP_WIDTH 28 /* when outputting, assume this width */
 #define FORMAT_TIMESTAMP_RELATIVE_MAX 256
 #define FORMAT_TIMESPAN_MAX 64