Service Desk (from tk@giga.or.at): gnutls-3.6.3: two problems on NetBSD

Hi!

There are two problems with gnutls-3.6.3 on NetBSD.

One is a compilation issue: NetBSD does not provide strdupa(), but this is used unconditionally. The attached patch changes the code to use the standard strdup() instead.

The second one is a core dump in a self test:

FAIL: p11-kit-trust.sh

patch-doc_examples_tlsproxy_tlsproxy.c

Thanks ! The strdupa() issue has already been addressed in the master branch.

made the issue visible to everyone

Can you provide more info on the core dump?

(gdb) bt #0 0x000076184023eb3a in kill () from /usr/lib/libc.so.12 #1 (closed) 0x0000761841a0b430 in pthread__assertfunc (file=file@entry=0x761841a0c3d0 "/usr/src/lib/libpthread/pthread_tsd.c", line=line@entry=176, function=function@entry=0x761841a0c480 "pthread__add_specific", expr=expr@entry=0x761841a0c410 "pthread__tsd_destructors[key] != NULL") at /usr/src/lib/libpthread/pthread.c:1100 #2 (closed) 0x0000761841a07aed in pthread__add_specific (self=0x761842f09800, key=0, value=value@entry=0x0) at /usr/src/lib/libpthread/pthread_tsd.c:176 #3 (closed) 0x0000761841a07d86 in pthread_setspecific (key=, value=value@entry=0x0) at /usr/src/lib/libpthread/pthread_specific.c:69 #4 0x00007618426f6546 in p11_library_uninit () at common/library.c:153 #5 (closed) 0x000076184262f509 in ?? () from /usr/pkg/lib/libp11-kit.so.0 #6 (closed) 0x0000761842f16d00 in ?? () #7 (closed) 0x00007618426f6629 in ?? () from /usr/pkg/lib/libp11-kit.so.0 #8 (closed) 0x0000000000000000 in ?? ()

Looking at the p11-kit source code I see:

void p11_library_uninit (void) { uninit_common ();

    /* Some cleanup to pacify valgrind */
    free (pthread_getspecific (thread_local));
    pthread_setspecific (thread_local, NULL);

#ifdef HAVE_STRERROR_L freelocale (p11_message_locale); #endif p11_message_storage = dont_store_message; pthread_key_delete (thread_local); p11_mutex_uninit (&p11_virtual_mutex); p11_mutex_uninit (&p11_library_mutex); }

where pthread_setspecific is in line 153.

Is this possibly a problem in p11-kit instead? Thomas

P.S.: Is this bug report available via HTTP(S) as well?

The issue looks p11-kit related, and there have been few changes in the latest versions regarding threading. You may want to try the latest version.

Yes, the report is at: #544 (closed)

Thank you for the pointer!

Updating to the latest p11-kit did indeed fix this issue.

There's a new error now, probably hidden before because tests stop when there are failures in one directory:

FAIL: pkcs12-utf8

For utf8 gnutls relies on libunistring (and in some cases libidn2). Do you have the latest version of these libs?

For utf8 gnutls relies on libunistring (and in some cases libidn2). Do you have the latest version of these libs?

libunistring 0.9.10 is installed, https://www.gnu.org/software/libunistring/#TOCdownloading says it's the latest.

libidn2 was not installed, but when I add it, the test fails in the same way. Thomas

pkcs12-utf8 provides utf8 on command line, could it be that the shell does not handle that? If you replace /bin/sh with /bin/bash does it change anything?

added 1 deleted label

Using bash in the script does indeed fix the problem! Thomas

mentioned in commit 55da5e6c

assigned to @nmav

mentioned in commit 7fbeabe5

mentioned in commit fefb3b43

closed via commit fefb3b43

closed via merge request !758 (merged)

removed milestone