Bug #64461 | assorted regression tests fail when built with openssl 1.0.1 | ||
---|---|---|---|
Submitted: | 27 Feb 2012 5:20 | Modified: | 21 Aug 2014 6:07 |
Reporter: | [ name withheld ] | Email Updates: | |
Status: | Closed | Impact on me: | |
Category: | MySQL Server: Tests | Severity: | S3 (Non-critical) |
Version: | 5.5.21, 5.5.23 | OS: | Any |
Assigned to: | CPU Architecture: | Any |
[27 Feb 2012 5:20]
[ name withheld ]
[27 Feb 2012 21:04]
Sveta Smirnova
Thank you for the report. Verified as described. Example of backtrace for last test: Thread 1 (Thread 0x7f357a10d720 (LWP 29389)): #0 0x0000003a4d00c6bc in pthread_kill () from /lib64/libpthread.so.0 #1 0x00000000008cf3ab in my_write_core (sig=11) at /home/sveta/src/mysql-5.5/mysys/stacktrace.c:433 #2 0x000000000075a8eb in handle_fatal_signal (sig=11) at /home/sveta/src/mysql-5.5/sql/signal_handler.cc:246 #3 <signal handler called> #4 0x00000031d75259d6 in __strcmp_sse42 () from /lib64/libc.so.6 #5 0x000000320bab451a in ?? () from /usr/lib64/libcrypto.so.10 #6 0x000000320bab4866 in lh_insert () from /usr/lib64/libcrypto.so.10 #7 0x000000320ba6071b in OBJ_NAME_add () from /usr/lib64/libcrypto.so.10 #8 0x00007f357a539af5 in SSL_library_init () from /usr/local/ssl/lib/libssl.so.1.0.0 #9 0x0000000000b6b251 in check_ssl_init () at /home/sveta/src/mysql-5.5/vio/viosslfactories.c:151 #10 0x0000000000b6b348 in new_VioSSLFd (key_file=0x155d9da "/home/sveta/src/mysql-5.5/mysql-test/std_data/server-key.pem", cert_file=0x155d98b "/home/sveta/src/mysql-5.5/mysql-test/std_data/server-cert.pem", ca_file=0x155d941 "/home/sveta/src/mysql-5.5/mysql-test/std_data/cacert.pem", ca_path=0x0, cipher=0x0, method=0x7f357a74bbe0, error=0x7fff19b6e3dc) at /home/sveta/src/mysql-5.5/vio/viosslfactories.c:182 #11 0x0000000000b6b770 in new_VioSSLAcceptorFd (key_file=0x155d9da "/home/sveta/src/mysql-5.5/mysql-test/std_data/server-key.pem", cert_file=0x155d98b "/home/sveta/src/mysql-5.5/mysql-test/std_data/server-cert.pem", ca_file=0x155d941 "/home/sveta/src/mysql-5.5/mysql-test/std_data/cacert.pem", ca_path=0x0, cipher=0x0, error=0x7fff19b6e3dc) at /home/sveta/src/mysql-5.5/vio/viosslfactories.c:286 #12 0x000000000055babc in init_ssl () at /home/sveta/src/mysql-5.5/sql/mysqld.cc:3596 #13 0x000000000055cf8d in mysqld_main (argc=50, argv=0x155da50) at /home/sveta/src/mysql-5.5/sql/mysqld.cc:4395 #14 0x00000000005558e4 in main (argc=8, argv=0x7fff19b6e568) at /home/sveta/src/mysql-5.5/sql/main.cc:25
[9 Mar 2012 13:55]
Honza Horak
As described above, we cannot rely on cipher name used in the tests, unless we specify the cipher explicitly. There are generally two situations where a cipher should be specified: 1) when executing a command using mysql utility in a test (see e.g. ssl_8k_key.test) - then we can use --ssl-cipher argument and specify cipher, that is expected by ssl_8k_key.result 2) when creating a new connection using 'connect' command in a test, which is sometimes used together with grants requiring specific cipher - see e.g. openssl_1.test. This is more tricky, because connect command accepts only general option SSL, but doesn't allow to specify a cipher currently. This means, that if we try to connect as a user, that is required to connect only using DHE-RSA-AES256-SHA, the test won't pass, because without specifying the cipher, the default value DHE-RSA-AES256-SHA256 will be used and it'll lead to connection refused. I'll attach a patch, that extends 'connect' command to accept "CIPHER:DHE-RSA-AES256-SHA" option, same as SSL/COMPRESS or other options. Using this we can enhance tests to use correct cipher explicitly, which is also used in the patch, so all ssl-related tests pass regardless of default cipher value. Please, consider using something like this patch, thanks.
[9 Mar 2012 13:57]
Honza Horak
proposed patch - specify the cipher explicitely
Attachment: mysql-cipherspec.patch (text/x-patch), 4.82 KiB.
[9 Mar 2012 14:12]
Honza Horak
proposed patch - specify the cipher explicitely (openssl_1.test fixed)
Attachment: mysql-cipherspec.patch (text/x-patch), 5.33 KiB.
[27 Mar 2012 6:26]
Valeriy Kravchuk
Bug #64773 was marked as a duplicate of this one.
[21 Aug 2014 6:07]
Murthy Sidagam
The tests were failing with openssl-1.0.1. These are passing with latest openssl-1.0.1h version build Hence, I feel there is no need to do any changes to test cases. And closing the bug