32bit limit for unix timestamp / OpenBSD

[hoppedmalt]

Describe the bug
32bit limit for unix timestamps on OpenBSD, Y2038 bug

To Reproduce
Prebuilt jq from OpenBSD ports tree

user@testhost:~/build/jq-1.7.1$ echo '"2038-01-19T03:14:07Z"' | jq 'fromdate'            
2147483647
user@testhost:~/build/jq-1.7.1$ echo '"2038-01-19T03:14:08Z"' | jq 'fromdate'            
-2147483648

Locally built jq from jq source package

user@testhost:~/build/jq-1.7.1$ echo '"2038-01-19T03:14:07Z"' | ./jq 'fromdate'            
2147483647
user@testhost:~/build/jq-1.7.1$ echo '"2038-01-19T03:14:08Z"' | ./jq 'fromdate'            
-2147483648

Expected behavior
Positive integer representation of seconds since epoch should be printed in both cases.

user@macos ~ %  echo '"2038-01-19T03:14:08Z"' | jq 'fromdate'  
2147483648
[user@linux ~]$  echo '"2038-01-19T03:14:08Z"' | jq 'fromdate'  
2147483648

Environment (please complete the following information):

• OS and Version: OpenBSD 7.6 amd64
• jq version 1.7.1

Additional context
Behavior of jq is consistent with both, distribution supplied jq package and jq built from source.
Looks like tests check with date "2015-03-05T23:51:47Z". Test with 2045?

[itchyny]

Isn't this an issue of platform library?

[hoppedmalt]

Possibly. Having a test in jq build for correct behavior should have more leverage than any bug report I can write, if that is the case.

[itchyny]

Can you run the following code?

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

int main () {
  struct tm tm;
  memset(&tm, 0, sizeof(tm));
  const char *end = strptime("2038-01-19T03:14:08Z", "%Y-%m-%dT%H:%M:%SZ", &tm);
  if (end == NULL) return 1;
  printf("%d-%02d-%02d %02d:%02d:%02d\n",
      tm.tm_year + 1900, tm.tm_mon + 1,
      tm.tm_mday, tm.tm_hour, tm.tm_min, tm.tm_sec);
}

[hoppedmalt]

$ ./a.out                                                                                  
2038-01-19 03:14:08

[itchyny]

printf("%ld\n", timegm(&tm)); prints 2147483648 or -2147483648?

[itchyny]

Or specifically sizeof(time_t) is 8 or 4?

[hoppedmalt]

8

[itchyny]

printf("%.0lf\n", (double) timegm(&tm)); is 2147483648?

[hoppedmalt]

2147483648, yes

[itchyny]

Hmm, confused.

[hoppedmalt]

Prototype introduction was missing in the default openbsd build of jq, adding this into builtin.c before use helped:

time_t timegm(struct tm *tm);

It seems that adding this definition into builtin.c's beginning helped to get the time.h read as well:

# define _BSD_SOURCE 1

Missing proto caused casting problems in my_mktime()

[itchyny]

Add to CFLAGS works?

diff --git a/Makefile.am b/Makefile.am
index c7504e4..cb09065 100644
--- a/Makefile.am
+++ b/Makefile.am
@@ -20,7 +20,7 @@ LIBJQ_SRC = src/builtin.c src/bytecode.c src/compile.c src/execute.c    \
 ### C build options
 
 AM_CFLAGS = -Wextra -Wall -Wno-unused-parameter -Wno-unused-function \
-            -Woverlength-strings
+            -Woverlength-strings -D_BSD_SOURCE -D_DEFAULT_SOURCE
 
 if WIN32
 AM_CFLAGS += -municode

[hoppedmalt]

Doesn't appear to work.