Hi,
I'm building 1.16.2 with just https://pagure.io/SSSD/sssd/c/a2cc554f438c220b3cc73eb93879dd87795a86cd?branc... applied (without it, it doesn't build in Ubuntu currently) and I'm seeing this test failure:
[==========] Running 2 test(s). [ RUN ] test_getsidbyname [ ERROR ] --- 0x2 != 0 [ LINE ] --- ../src/tests/cmocka/sss_nss_idmap-tests.c:121: error: Failure! [ FAILED ] test_getsidbyname [ RUN ] test_getorigbyname [ ERROR ] --- 0x2 != 0 [ LINE ] --- ../src/tests/cmocka/sss_nss_idmap-tests.c:140: error: Failure! [ FAILED ] test_getorigbyname [==========] 2 test(s) run. [ PASSED ] 0 test(s). [ FAILED ] 2 test(s), listed below: [ FAILED ] test_getsidbyname [ FAILED ] test_getorigbyname
2 FAILED TEST(S) FAIL sss_nss_idmap-tests (exit status: 2)
I tried with samba 4.7.6 and 4.8.2 installed, and also with --with-smb-idmap-interface-version 5 and 6, same result. Debian is at 1.16.2 and the tests pass there just fine, so I think I'm looking at some dependency problem. ldb is 1.3.1 tdb is 1.3.15
Any pointers? Maybe a way to run just that test, so I can add debugging statements?
Thanks!
What I figured out so far is that this is a test that is enabled if you have cmocka installed, and this is the first time I had that. On Fri, Jul 20, 2018 at 2:22 PM Andreas Hasenack andreas@canonical.com wrote:
Hi,
I'm building 1.16.2 with just https://pagure.io/SSSD/sssd/c/a2cc554f438c220b3cc73eb93879dd87795a86cd?branc... applied (without it, it doesn't build in Ubuntu currently) and I'm seeing this test failure:
[==========] Running 2 test(s). [ RUN ] test_getsidbyname [ ERROR ] --- 0x2 != 0 [ LINE ] --- ../src/tests/cmocka/sss_nss_idmap-tests.c:121: error: Failure! [ FAILED ] test_getsidbyname [ RUN ] test_getorigbyname [ ERROR ] --- 0x2 != 0 [ LINE ] --- ../src/tests/cmocka/sss_nss_idmap-tests.c:140: error: Failure! [ FAILED ] test_getorigbyname [==========] 2 test(s) run. [ PASSED ] 0 test(s). [ FAILED ] 2 test(s), listed below: [ FAILED ] test_getsidbyname [ FAILED ] test_getorigbyname
2 FAILED TEST(S) FAIL sss_nss_idmap-tests (exit status: 2)
I tried with samba 4.7.6 and 4.8.2 installed, and also with --with-smb-idmap-interface-version 5 and 6, same result. Debian is at 1.16.2 and the tests pass there just fine, so I think I'm looking at some dependency problem. ldb is 1.3.1 tdb is 1.3.15
Any pointers? Maybe a way to run just that test, so I can add debugging statements?
Thanks!
Unfortunately these tests don’t have an option to raise the debug level so stepping throught them with gdb is the only option I’m afraid..
On 20 Jul 2018, at 20:56, Andreas Hasenack andreas@canonical.com wrote:
What I figured out so far is that this is a test that is enabled if you have cmocka installed, and this is the first time I had that. On Fri, Jul 20, 2018 at 2:22 PM Andreas Hasenack andreas@canonical.com wrote:
Hi,
I'm building 1.16.2 with just https://pagure.io/SSSD/sssd/c/a2cc554f438c220b3cc73eb93879dd87795a86cd?branc... applied (without it, it doesn't build in Ubuntu currently) and I'm seeing this test failure:
[==========] Running 2 test(s). [ RUN ] test_getsidbyname [ ERROR ] --- 0x2 != 0 [ LINE ] --- ../src/tests/cmocka/sss_nss_idmap-tests.c:121: error: Failure! [ FAILED ] test_getsidbyname [ RUN ] test_getorigbyname [ ERROR ] --- 0x2 != 0 [ LINE ] --- ../src/tests/cmocka/sss_nss_idmap-tests.c:140: error: Failure! [ FAILED ] test_getorigbyname [==========] 2 test(s) run. [ PASSED ] 0 test(s). [ FAILED ] 2 test(s), listed below: [ FAILED ] test_getsidbyname [ FAILED ] test_getorigbyname
2 FAILED TEST(S) FAIL sss_nss_idmap-tests (exit status: 2)
I tried with samba 4.7.6 and 4.8.2 installed, and also with --with-smb-idmap-interface-version 5 and 6, same result. Debian is at 1.16.2 and the tests pass there just fine, so I think I'm looking at some dependency problem. ldb is 1.3.1 tdb is 1.3.15
Any pointers? Maybe a way to run just that test, so I can add debugging statements?
Thanks!
sssd-users mailing list -- sssd-users@lists.fedorahosted.org To unsubscribe send an email to sssd-users-leave@lists.fedorahosted.org Fedora Code of Conduct: https://getfedora.org/code-of-conduct.html List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/sssd-users@lists.fedorahosted....
On Mon, Jul 23, 2018 at 10:01:26AM +0200, Jakub Hrozek wrote:
Unfortunately these tests don’t have an option to raise the debug level so stepping throught them with gdb is the only option I’m afraid..
I think I didn't properly mock sss_nss_make_request_timeout() here. Instead of the provided call which just mock the results the original one is used which tries to talk to SSSD which either does not run or does not know about the test user, hence the return code 0x02 (ENOENT).
If you run the test with strace you should see that the test program tries to connect to /var/lib/sss/pipes/nss which is not expected. I'll try to fix this.
bye, Sumit
On 20 Jul 2018, at 20:56, Andreas Hasenack andreas@canonical.com wrote:
What I figured out so far is that this is a test that is enabled if you have cmocka installed, and this is the first time I had that. On Fri, Jul 20, 2018 at 2:22 PM Andreas Hasenack andreas@canonical.com wrote:
Hi,
I'm building 1.16.2 with just https://pagure.io/SSSD/sssd/c/a2cc554f438c220b3cc73eb93879dd87795a86cd?branc... applied (without it, it doesn't build in Ubuntu currently) and I'm seeing this test failure:
[==========] Running 2 test(s). [ RUN ] test_getsidbyname [ ERROR ] --- 0x2 != 0 [ LINE ] --- ../src/tests/cmocka/sss_nss_idmap-tests.c:121: error: Failure! [ FAILED ] test_getsidbyname [ RUN ] test_getorigbyname [ ERROR ] --- 0x2 != 0 [ LINE ] --- ../src/tests/cmocka/sss_nss_idmap-tests.c:140: error: Failure! [ FAILED ] test_getorigbyname [==========] 2 test(s) run. [ PASSED ] 0 test(s). [ FAILED ] 2 test(s), listed below: [ FAILED ] test_getsidbyname [ FAILED ] test_getorigbyname
2 FAILED TEST(S) FAIL sss_nss_idmap-tests (exit status: 2)
I tried with samba 4.7.6 and 4.8.2 installed, and also with --with-smb-idmap-interface-version 5 and 6, same result. Debian is at 1.16.2 and the tests pass there just fine, so I think I'm looking at some dependency problem. ldb is 1.3.1 tdb is 1.3.15
Any pointers? Maybe a way to run just that test, so I can add debugging statements?
Thanks!
sssd-users mailing list -- sssd-users@lists.fedorahosted.org To unsubscribe send an email to sssd-users-leave@lists.fedorahosted.org Fedora Code of Conduct: https://getfedora.org/code-of-conduct.html List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/sssd-users@lists.fedorahosted....
sssd-users mailing list -- sssd-users@lists.fedorahosted.org To unsubscribe send an email to sssd-users-leave@lists.fedorahosted.org Fedora Code of Conduct: https://getfedora.org/code-of-conduct.html List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/sssd-users@lists.fedorahosted....
On Tue, Aug 7, 2018 at 8:04 AM Sumit Bose sbose@redhat.com wrote:
On Mon, Jul 23, 2018 at 10:01:26AM +0200, Jakub Hrozek wrote:
Unfortunately these tests don’t have an option to raise the debug level so stepping throught them with gdb is the only option I’m afraid..
I think I didn't properly mock sss_nss_make_request_timeout() here. Instead of the provided call which just mock the results the original one is used which tries to talk to SSSD which either does not run or does not know about the test user, hence the return code 0x02 (ENOENT).
If you run the test with strace you should see that the test program tries to connect to /var/lib/sss/pipes/nss which is not expected. I'll try to fix this.
Indeed it does try that connect a few times:
11933 connect(3, {sa_family=AF_UNIX, sun_path="/var/lib/sss/pipes/nss"}, 110) = -1 ENOENT (No such file or directory)
But something is still unexplained: the same test works just fine in debian, and doesn't try to connect to that socket.
I might try updating nss. I have 3.36, and debian has 3.38.
Thanks!
On Tue, Aug 07, 2018 at 09:36:25AM -0300, Andreas Hasenack wrote:
On Tue, Aug 7, 2018 at 8:04 AM Sumit Bose sbose@redhat.com wrote:
On Mon, Jul 23, 2018 at 10:01:26AM +0200, Jakub Hrozek wrote:
Unfortunately these tests don’t have an option to raise the debug level so stepping throught them with gdb is the only option I’m afraid..
I think I didn't properly mock sss_nss_make_request_timeout() here. Instead of the provided call which just mock the results the original one is used which tries to talk to SSSD which either does not run or does not know about the test user, hence the return code 0x02 (ENOENT).
If you run the test with strace you should see that the test program tries to connect to /var/lib/sss/pipes/nss which is not expected. I'll try to fix this.
Indeed it does try that connect a few times:
11933 connect(3, {sa_family=AF_UNIX, sun_path="/var/lib/sss/pipes/nss"}, 110) = -1 ENOENT (No such file or directory)
Thanks for the confirmation.
But something is still unexplained: the same test works just fine in debian, and doesn't try to connect to that socket.
This is just linker magic. Due to my fault sss_nss_make_request_timeout() is defined twice and which symbol is picked might depend on specific linker options used.
bye, Sumit
I might try updating nss. I have 3.36, and debian has 3.38.
Thanks! _______________________________________________ sssd-users mailing list -- sssd-users@lists.fedorahosted.org To unsubscribe send an email to sssd-users-leave@lists.fedorahosted.org Fedora Code of Conduct: https://getfedora.org/code-of-conduct.html List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/sssd-users@lists.fedorahosted....
On Tue, Aug 7, 2018 at 10:19 AM Sumit Bose sbose@redhat.com wrote:
But something is still unexplained: the same test works just fine in debian, and doesn't try to connect to that socket.
This is just linker magic. Due to my fault sss_nss_make_request_timeout() is defined twice and which symbol is picked might depend on specific linker options used.
I just found out what option that was. Ubuntu, since many years, uses -Wl,-Bsymbolic-functions in its default linker flags. Debian doesn't.
The moment I strip this flag from the build, the test passes. I can't do that for the official package build, but it's good to know what option was causing the test to fail.
On (07/08/18 15:48), Andreas Hasenack wrote:
On Tue, Aug 7, 2018 at 10:19 AM Sumit Bose sbose@redhat.com wrote:
But something is still unexplained: the same test works just fine in debian, and doesn't try to connect to that socket.
This is just linker magic. Due to my fault sss_nss_make_request_timeout() is defined twice and which symbol is picked might depend on specific linker options used.
I just found out what option that was. Ubuntu, since many years, uses -Wl,-Bsymbolic-functions in its default linker flags. Debian doesn't.
The moment I strip this flag from the build, the test passes. I can't do that for the official package build, but it's good to know what option was causing the test to fail.
That would need to be disabled just for make check. Because libsss_nss_idmap_tests.so which is used for testing is not the same as libsss_nss_idmap.so which is used in reality.
libsss_nss_idmap_tests.so also export function sss_nss_make_request_timeout which is not exported in libsss_nss_idmap.so
sh-4.4$ nm --defined-only --dynamic .libs/libsss_nss_idmap.so | grep request sh-4.4$ nm --defined-only --dynamic .libs/libsss_nss_idmap_tests.so | grep request 0000000000004290 T sss_nss_make_request_timeout
It was done to make testing possible and avoid copy&paste mistakes
libsss_nss_idmap_tests_la_SOURCES = $(libsss_nss_idmap_la_SOURCES) libsss_nss_idmap_tests_la_LIBADD = $(libsss_nss_idmap_la_LIBADD) libsss_nss_idmap_tests_la_LDFLAGS = \ $(libsss_nss_idmap_la_LDFLAGS) \ -shared \ -rpath $(libdir) \ -Wl,--version-script,$(srcdir)/src/sss_client/idmap/sss_nss_idmap.unit_tests
Other option would be to "include" all sources from library to test. + use -Wl,-wrap,sss_nss_make_request_timeout
Or another hacky way to use weak symbols
diff --git a/Makefile.am b/Makefile.am index ea7648bcd..7216561fb 100644 --- a/Makefile.am +++ b/Makefile.am @@ -2600,12 +2600,13 @@ test_authtok_LDADD = \ $(NULL)
sss_nss_idmap_tests_SOURCES = \ - src/tests/cmocka/sss_nss_idmap-tests.c + src/tests/cmocka/sss_nss_idmap-tests.c \ + $(libsss_nss_idmap_la_SOURCES) sss_nss_idmap_tests_CFLAGS = \ $(AM_CFLAGS) sss_nss_idmap_tests_LDADD = \ $(CMOCKA_LIBS) \ - libsss_nss_idmap_tests.la \ + $(libsss_nss_idmap_la_LIBADD) \ $(NULL)
deskprofile_utils_tests_SOURCES = \ diff --git a/src/sss_client/common.c b/src/sss_client/common.c index 67a460705..a93aaff16 100644 --- a/src/sss_client/common.c +++ b/src/sss_client/common.c @@ -718,6 +718,7 @@ static enum sss_status sss_cli_check_socket(int *errnop,
/* this function will check command codes match and returned length is ok */ /* repbuf and replen report only the data section not the header */ +__attribute((weak)) enum nss_status sss_nss_make_request_timeout(enum sss_cli_command cmd, struct sss_cli_req_data *rd, int timeout,
LS
On Tue, Aug 07, 2018 at 10:38:59PM +0200, Lukas Slebodnik wrote:
On (07/08/18 15:48), Andreas Hasenack wrote:
On Tue, Aug 7, 2018 at 10:19 AM Sumit Bose sbose@redhat.com wrote:
But something is still unexplained: the same test works just fine in debian, and doesn't try to connect to that socket.
This is just linker magic. Due to my fault sss_nss_make_request_timeout() is defined twice and which symbol is picked might depend on specific linker options used.
I just found out what option that was. Ubuntu, since many years, uses -Wl,-Bsymbolic-functions in its default linker flags. Debian doesn't.
The moment I strip this flag from the build, the test passes. I can't do that for the official package build, but it's good to know what option was causing the test to fail.
That would need to be disabled just for make check. Because libsss_nss_idmap_tests.so which is used for testing is not the same as libsss_nss_idmap.so which is used in reality.
libsss_nss_idmap_tests.so also export function sss_nss_make_request_timeout which is not exported in libsss_nss_idmap.so
sh-4.4$ nm --defined-only --dynamic .libs/libsss_nss_idmap.so | grep request sh-4.4$ nm --defined-only --dynamic .libs/libsss_nss_idmap_tests.so | grep request 0000000000004290 T sss_nss_make_request_timeout
It was done to make testing possible and avoid copy&paste mistakes
libsss_nss_idmap_tests_la_SOURCES = $(libsss_nss_idmap_la_SOURCES) libsss_nss_idmap_tests_la_LIBADD = $(libsss_nss_idmap_la_LIBADD) libsss_nss_idmap_tests_la_LDFLAGS = \ $(libsss_nss_idmap_la_LDFLAGS) \ -shared \ -rpath $(libdir) \ -Wl,--version-script,$(srcdir)/src/sss_client/idmap/sss_nss_idmap.unit_tests
Other option would be to "include" all sources from library to test.
- use -Wl,-wrap,sss_nss_make_request_timeout
Or another hacky way to use weak symbols
Thank you for figuring out the linker option which caused the issue and for the suggestions.
I've opened https://pagure.io/SSSD/sssd/issue/3801 to track the issue and also created https://github.com/SSSD/sssd/pull/632.
I preferred using -Wl,-wrap to make it more consistent with other tests and hopefully also easier to understand.
bye, Sumit
diff --git a/Makefile.am b/Makefile.am index ea7648bcd..7216561fb 100644 --- a/Makefile.am +++ b/Makefile.am @@ -2600,12 +2600,13 @@ test_authtok_LDADD = \ $(NULL)
sss_nss_idmap_tests_SOURCES = \
- src/tests/cmocka/sss_nss_idmap-tests.c
- src/tests/cmocka/sss_nss_idmap-tests.c \
- $(libsss_nss_idmap_la_SOURCES)
sss_nss_idmap_tests_CFLAGS = \ $(AM_CFLAGS) sss_nss_idmap_tests_LDADD = \ $(CMOCKA_LIBS) \
- libsss_nss_idmap_tests.la \
- $(libsss_nss_idmap_la_LIBADD) \ $(NULL)
deskprofile_utils_tests_SOURCES = \ diff --git a/src/sss_client/common.c b/src/sss_client/common.c index 67a460705..a93aaff16 100644 --- a/src/sss_client/common.c +++ b/src/sss_client/common.c @@ -718,6 +718,7 @@ static enum sss_status sss_cli_check_socket(int *errnop,
/* this function will check command codes match and returned length is ok */ /* repbuf and replen report only the data section not the header */ +__attribute((weak)) enum nss_status sss_nss_make_request_timeout(enum sss_cli_command cmd, struct sss_cli_req_data *rd, int timeout,
LS
sssd-users mailing list -- sssd-users@lists.fedorahosted.org To unsubscribe send an email to sssd-users-leave@lists.fedorahosted.org Fedora Code of Conduct: https://getfedora.org/code-of-conduct.html List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/sssd-users@lists.fedorahosted....
Thank you for figuring out the linker option which caused the issue and for the suggestions.
I've opened https://pagure.io/SSSD/sssd/issue/3801 to track the issue and also created https://github.com/SSSD/sssd/pull/632.
Thanks. I commented in the PR. The test now passes on Ubuntu with -Wl,-Bsymbolic-functions enabled. \o/
sssd-users@lists.fedorahosted.org
-
Andreas Hasenack -
Jakub Hrozek -
Lukas Slebodnik -
Sumit Bose