got fails to build: Undefined symbols: "_memset_s"

[barracuda156]

Undefined symbols:
  "_memset_s", referenced from:
      _recallocarray in libopenbsd-compat.a(recallocarray.o)
ld: symbol(s) not found
collect2: ld returned 1 exit status
make[2]: *** [got-fetch-pack] Error 1
make[2]: Leaving directory `/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_macports_release_tarballs_ports_devel_got/got/work/got-portable-0.77/libexec/got-fetch-pack'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_macports_release_tarballs_ports_devel_got/got/work/got-portable-0.77/libexec'
make: *** [all-recursive] Error 1

Same failure is confirmed on 10.6.8 Rosetta (gcc-4.2, gcc12) and 10.8.5 (clang, gcc12).

[barracuda156]

Adding -D__STDC_WANT_LIB_EXT1__=1 to cppflags or cflags fails to help.

[artkiver (グレェ)]

Hi @baracuda156 thanks for the bug report.

To be honest, I am uncertain if I will be able to be of much help on this. While I do have some Leopard vintage G4 PowerBooks in storage, I am not sure if I have anything older than perhaps Catalina running on any of my personal systems easily accessible? Moreover, since the port for gameoftrees was created, I don't think the details view (e.g. ​https://ports.macports.org/port/got/details/) has ever shown got building successfully on any version of the OS older than High Sierra.

In other words: this may be a more challenging bug to address than I personally can solve. However, if you or others can find any workable patches, the upstream project is pretty cooperative in getting such things committed.

[artkiver (グレェ)]

Known working and not working got OS versions.

[barracuda156]

Replying to artkiver:

Thank you for responding. I will try fixing it.

[kencu (Ken)]

memset_s was not available until MacOSX 10.9:

errno_t	memset_s(void *__s, rsize_t __smax, int __c, rsize_t __n) __OSX_AVAILABLE_STARTING(__MAC_10_9, __IPHONE_7_0);

but it does not look hard to write a similar function, so it could easily be put into legacysupport.

​https://github.com/apple/darwin-xnu/blob/main/osfmk/kern/memset_s.c

[artkiver (グレェ)]

@barracuda156 You are most welcome. I would say as far as other resources, Thomas Adam is the got-portable maintainer, albeit as far as I know, he mostly focuses on Linux build issues but was still cooperative with me when -portable had an issue introduced after adding a diff for compatibility with Homebrew, which caused challenges with the MacPorts assumptions. There is also an IRC channel #gameoftrees on libera.chat and a low volume mailing list (more information on that here: ​https://lists.openbsd.org/cgi-bin/mj_wwwusr?user=&passw=&func=lists-long-full&extra=gameoftrees ) which may be helpful?

@kencu if that methodology is viable, that would be great! Thanks for the extra set of eyeballs.

[kencu (Ken)]

the other way to approach this is for upstream to test for and only use memset_s if the symbol exists, otherwise fall back (presumably) to memset.

[barracuda156]

Replying to kencu:

the other way to approach this is for upstream to test for and only use memset_s if the symbol exists, otherwise fall back (presumably) to memset.

samba4 has this:

#ifndef HAVE_MEMSET_S
# ifndef RSIZE_MAX
#  define RSIZE_MAX (SIZE_MAX >> 1)
# endif

int rep_memset_s(void *dest, size_t destsz, int ch, size_t count)
{
	if (dest == NULL) {
		return EINVAL;
	}

	if (destsz > RSIZE_MAX ||
	    count > RSIZE_MAX ||
	    count > destsz) {
		return ERANGE;
	}

#if defined(HAVE_MEMSET_EXPLICIT)
	memset_explicit(dest, destsz, ch, count);
#else /* HAVE_MEMSET_EXPLICIT */
	memset(dest, ch, count);
# if defined(HAVE_GCC_VOLATILE_MEMORY_PROTECTION)
	/* See http://llvm.org/bugs/show_bug.cgi?id=15495 */
	__asm__ volatile("" : : "g"(dest) : "memory");
# endif /* HAVE_GCC_VOLATILE_MEMORY_PROTECTION */
#endif /* HAVE_MEMSET_EXPLICIT */

	return 0;
}
#endif /* HAVE_MEMSET_S */

And postgresql has:

#if defined(HAVE_MEMSET_S)

void
explicit_bzero(void *buf, size_t len)
{
	(void) memset_s(buf, len, 0, len);
}

#elif defined(WIN32)

void
explicit_bzero(void *buf, size_t len)
{
	(void) SecureZeroMemory(buf, len);
}

#else

/*
 * Indirect call through a volatile pointer to hopefully avoid dead-store
 * optimisation eliminating the call.  (Idea taken from OpenSSH.)  We can't
 * assume bzero() is present either, so for simplicity we define our own.
 */

static void
bzero2(void *buf, size_t len)
{
	memset(buf, 0, len);
}

static void (*volatile bzero_p) (void *, size_t) = bzero2;

void
explicit_bzero(void *buf, size_t len)
{
	bzero_p(buf, len);
}

#endif

Anything is this usable? A check for HAVE_MEMSET_S is easily added to configure.

[artkiver (グレェ)]

As a bit of an aside, got 0.78 was released and I am prepping a PR for that. It won't address the issues raised in this ticket, but I thought I would keep you apprised.

[artkiver (グレェ)]

Still somewhat of an aside, I submitted a PR for got 0.79 and that was merged. I was also looking into some other tickets, and there was mention of legacy support devel (​https://ports.macports.org/port/legacy-support-devel/) perhaps that might be a good place to introduce the HAVE_MEMSET_S function and test it on older OSes to see if that may help with got?

[barracuda156]

Replying to artkiver:

Still somewhat of an aside, I submitted a PR for got 0.79 and that was merged. I was also looking into some other tickets, and there was mention of legacy support devel (​https://ports.macports.org/port/legacy-support-devel/) perhaps that might be a good place to introduce the HAVE_MEMSET_S function and test it on older OSes to see if that may help with got?

Do you have a suggestion how to better write it?

[barracuda156]

Replying to kencu:

but it does not look hard to write a similar function, so it could easily be put into legacysupport.

​https://github.com/apple/darwin-xnu/blob/main/osfmk/kern/memset_s.c

Will this work?

​https://github.com/regehr/sudo-1.8.13/blob/master/lib/util/memset_s.c

[artkiver (グレェ)]

Possibly! To be honest, I never got around to digging up any older laptops from my storage unit to test even the earlier suggestions about augmenting legacy-support or legacy-support-devel and had kind of forgotten about it until the Trac update reminded me. I just reached out to Thomas Adams (who heads up the -portable branch of Got) to see if maybe a better approach would be to address this upstream in the event that there may be people running older Mac OS versions who don't want to run MacPorts.

Replying to barracuda156:

Replying to kencu:

but it does not look hard to write a similar function, so it could easily be put into legacysupport.

​https://github.com/apple/darwin-xnu/blob/main/osfmk/kern/memset_s.c

Will this work?

​https://github.com/regehr/sudo-1.8.13/blob/master/lib/util/memset_s.c

[barracuda156]

Replying to artkiver:

I just reached out to Thomas Adams (who heads up the -portable branch of Got) to see if maybe a better approach would be to address this upstream in the event that there may be people running older Mac OS versions who don't want to run MacPorts.

Thank you!

[aeiouaeiouaeiouaeiouaeiouaeiou]

In 59131f43d06df900d3288b9e9ec5a8bd5c3430a4/macports-ports (master):

got: fix build on macOS <10.9

• set legacysupport version

Closes: #66087