musl - Re: Re: freeswitch and musl 1.2.x (time64 most likely)

Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Fri, 29 Oct 2021 21:00:43 -0400
From: Rich Felker <dalias@...c.org>
To: musl@...ts.openwall.com, Szabolcs Nagy <nsz@...t70.net>,
	Érico Nogueira <ericonr@...root.org>
Subject: Re: Re: freeswitch and musl 1.2.x (time64 most likely)

On Sat, Oct 30, 2021 at 12:44:05AM +0200, Sebastian Kemper wrote:
> Am Fri, Oct 29, 2021 at 11:04:45PM +0200 schrieb Szabolcs Nagy:
> > * Sebastian Kemper <sebastian_ml@....net> [2021-10-29 16:40:02 +0200]:
> > > (gdb) p current
> > > $1 = 193277741781
> > > (gdb) p &current
> > > $2 = (apr_time_t *) 0x7fffac90
> > > (gdb) p os_now
> > > $3 = 19521141760827868
> > > (gdb) p &os_now
> > > $4 = (time_t *) 0x7fffac88
> > > (gdb) p timediff
> > > $5 = 19521141756395521
> >
> > these values are completely wrong (may be not set up yet,
> > you cannot rely on line numbers in optimized code)
> >
> > time on your system now should be around
> >
> > 1635500000
> >
> > current / 1000000 is
> >
> > 193277
> >
> > which is 1970-01-03 05:41:17
> >
> > os_now is far in the future.
> >
> 
> Hello Szabolcs,
> 
> I changed the test program a little bit:
> 
> --- a/test/testtime.c
> +++ b/test/testtime.c
> @@ -21,6 +21,7 @@
>  #include "testutil.h"
>  #include "apr_strings.h"
>  #include <time.h>
> +#include <stdio.h>
> 
>  #define STR_SIZE 45
> 
> @@ -57,6 +58,7 @@ static void test_now(abts_case *tc, void
>      apr_time_t timediff;
>      apr_time_t current;
>      time_t os_now;
> +    struct tm * timeinfo;
> 
>      current = apr_time_now();
>      time(&os_now);
> @@ -66,6 +68,10 @@ static void test_now(abts_case *tc, void
>       * that the time will be slightly off, so accept anything between -1 and
>       * 1 second.
>       */
> +    timeinfo = localtime ( &os_now );
> +    printf ( "Current local time and date: %s - %lld seconds since epoch\n", asctime (timeinfo), os_now);
> +    timeinfo = localtime ( &current );
> +    printf ( "Current APR time and date: %s - %lld seconds since epoch\n", asctime (timeinfo), current);
>      ABTS_ASSERT(tc, "apr_time and OS time do not agree",
>               (timediff > -2) && (timediff < 2));
>  }
> 
> The output is:
> 
> Current local time and date: Sat Oct 30 00:25:11 2021 - 7024617916842658549 seconds since epoch
> Current APR time and date: Mon Jan  5 10:46:45 1970 - 1635546302107561 seconds since epoch
> |Line 76: apr_time and OS time do not agree
> 
> (gdb) b time
> Breakpoint 1 at 0x77fd0b08: file compat/time32/time32.c, line 7.
> (gdb) b gettimeofday
> Breakpoint 2 at 0x77fcff6c: file compat/time32/gettimeofday_time32.c, line 7.
> (gdb) b apr_time_now
> Function "apr_time_now" not defined.
> Make breakpoint pending on future shared library load? (y or [n]) y
> Breakpoint 3 (apr_time_now) pending.
> (gdb) c
> Continuing.
> 
> Breakpoint 1, time (p=0x7fffac90) at compat/time32/time32.c:7
> 7	{
> (gdb) next
> 8		time_t t = time(0);
> (gdb)
> 9		if (t < INT32_MIN || t > INT32_MAX) {
> (gdb)
> 13		if (p) *p = t;
> (gdb)
> 14		return t;
> (gdb)
> test_now (tc=0x7fffacd0, data=0x0) at testtime.c:66
> 66	    timediff = os_now - (current / APR_USEC_PER_SEC);
> (gdb) p os_now
> $1 = 7024617916842658549
> (gdb) p current
> $2 = 1635546302107561
> (gdb) p APR_USEC_PER_SEC
> $3 = 1000000
> (gdb) c
> Continuing.
> [Inferior 1 (process 15506) exited with code 01]

This function should not be being called. It's only for ABI-compat
with old time32 binaries.

> So OS time is correct. Something not working right with apr.

Likely it is doing something bad bypassing the system headers and
declaring gettimeofday (wrongly) itself...

Rich
Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.