✘64-bit time_t on glibc 2.34 and up

[Gary E. Miller]

Yo Greg!

On Sat, 14 Jan 2023 09:32:03 -0500
Greg Troxel <gdt at lexort.com> wrote:

> "Gary E. Miller" <gem at rellim.com> writes:
> 
> >> Which magically changes references to syscalls that use time_t, in
> >> the entire binary, to the new ones?  
> >
> > Uh, once again, no.  No "versioning".  No syscalls are changed.
> >
> > Every existing syscall that uses 32-bit time_t now has a 64_bit
> > twin.  
> 
> And the new ones have new codepoints?

Yes.

> New names that are user-facing?

No.

> Does one have to ifdef _TIME_SIZE_BITS in source code and call
> gettimeofday64?

No.

> Or in a program with _TIME_SIZE_BITS there is some #define so that
> gettimeofday in the sources maps to the int64_t version?

Yes.  From /usr/include/sys/time.h:

#ifndef __USE_TIME_BITS64
extern int gettimeofday (struct timeval *__restrict __tv,
                         void *__restrict __tz) __THROW __nonnull ((1));
#else
# ifdef __REDIRECT_NTH
extern int __REDIRECT_NTH (gettimeofday, (struct timeval *__restrict __tv,
                                          void *__restrict __tz),
                           __gettimeofday64) __nonnull ((1));
# else
#  define gettimeofday __gettimeofday64
# endif
#endif


> This is what has been unclear

Maybe to BSD people...

> > So old and new binaries just work.  
> 
> That has to mean that the old syscall number leads to a syscall with
> the old behavior and the new syscall has a new number.  That's what I
> meant by versioned.

That is not how I use "versioned".  In Linux land "versioned" applis to the
version number on shared libraries.

> >> What happens if a library defines D_TIME_BITS 64 and makes
> >> syscalls, and a program which is unaware of this defines 32 and
> >> also makes sycalls? Or is the syscall switch per .o because the
> >> names are #defined?  
> >
> > That can never happen.  
> 
> I don't see how different .o files are prohibited from having
> different _TIME_SIZE_BITS, unless it leads to a link failure.

You can always build bad .o.  But good practice is to build all .o
in a program or binary with the same setting.  Otherwise a huge number
of ways to fail.  That is why all gpsd c files start with:

#include "../include/gpsd_config.h"  // must be before all includes

> Are there then new and old libc?

Of course. glibs versions change with the days of the week.  When a
program is linked, it is matched to one and only one glibc version.

libc proliferate like bunnies.  Thus the need for them to be versioned.

RGDS
GARY
---------------------------------------------------------------------------
Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97703
	gem at rellim.com  Tel:+1 541 382 8588

	    Veritas liberabit vos. -- Quid est veritas?
    "If you can't measure it, you can't improve it." - Lord Kelvin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 851 bytes
Desc: OpenPGP digital signature
URL: <https://lists.ntpsec.org/pipermail/devel/attachments/20230114/a2f4ec15/attachment.bin>