Details
-
Type:
Bug
-
Status: Won't Fix
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: mariadb
-
Labels:
-
Team:Data Access and Database
Description
I have a build failure in daf_persistence due to mariadbclient not linking to libssl/libcrypto on a linux machine. Can we make the ssloptions fix for OSX apply generically, or would that break something else?
Here's the daf_persistence error that I received:
daf_persistence: 2016_01.0-1-gf47bb69 ERROR (33 sec).
|
*** error building product daf_persistence.
|
*** exit code = 2
|
*** log is in /astro/apps6/opt/lsstStacks/lsstsw/build/daf_persistence/_build.log
|
*** last few lines:
|
::::: [2016-02-17T23:03:47.312613Z] /astro/apps6/opt/lsstStacks/lsstsw/stack/Linux64/mariadbclient/10.1.11/lib/libmysqlclient_r.so: undefined reference to `SSL_CTX_ctrl'
|
::::: [2016-02-17T23:03:47.312663Z] /astro/apps6/opt/lsstStacks/lsstsw/stack/Linux64/mariadbclient/10.1.11/lib/libmysqlclient_r.so: undefined reference to `ERR_get_error'
|
::::: [2016-02-17T23:03:47.312765Z] /astro/apps6/opt/lsstStacks/lsstsw/stack/Linux64/mariadbclient/10.1.11/lib/libmysqlclient_r.so: undefined reference to `SSL_new'
|
::::: [2016-02-17T23:03:47.312815Z] collect2: error: ld returned 1 exit status
|
::::: [2016-02-17T23:03:47.312834Z] scons: *** [tests/DbStorage_1] Error 1
|
::::: [2016-02-17T23:03:47.735155Z] /astro/apps6/opt/lsstStacks/lsstsw/miniconda/lib/python2.7/config/libpython2.7.a(posixmodule.o): In function `posix_tmpnam':
|
::::: [2016-02-17T23:03:47.735289Z] -------src-dir-------/Python-2.7.11/./Modules/posixmodule.c:7631: warning: the use of `tmpnam_r' is dangerous, better use `mkstemp'
|
::::: [2016-02-17T23:03:47.735345Z] /astro/apps6/opt/lsstStacks/lsstsw/miniconda/lib/python2.7/config/libpython2.7.a(posixmodule.o): In function `posix_tempnam':
|
::::: [2016-02-17T23:03:47.735400Z] -------src-dir-------/Python-2.7.11/./Modules/posixmodule.c:7578: warning: the use of `tempnam' is dangerous, better use `mkstemp'
|
::::: [2016-02-17T23:03:47.764830Z] scons: building terminated because of errors.
|
Here's what my mariadb lib linkage looks like, with the missing libs:
[16:32:53 parejkoj@magneto: /astro/apps6/opt/lsstStacks/lsstsw/stack/Linux64/mariadbclient/10.1.11/lib]
$ ldd libmysqlclient.so.18.0.0
linux-vdso.so.1 => (0x00007ffdb63b4000)
libpthread.so.0 => /lib64/libpthread.so.0 (0x00007f088ab10000)
libz.so.1 => /lib64/libz.so.1 (0x00007f088a8f9000)
libssl.so.1.0.0 => not found
libcrypto.so.1.0.0 => not found
libdl.so.2 => /lib64/libdl.so.2 (0x00007f088a6f4000)
librt.so.1 => /lib64/librt.so.1 (0x00007f088a4ec000)
libstdc+.so.6 => /usr/lib64/libstdc+.so.6 (0x00007f088a1e6000)
libm.so.6 => /lib64/libm.so.6 (0x00007f0889f61000)
libgcc_s.so.1 => /lib64/libgcc_s.so.1 (0x00007f0889d4b000)
libc.so.6 => /lib64/libc.so.6 (0x00007f08899b7000)
/lib64/ld-linux-x86-64.so.2 (0x00007f088b2af000)
Attachments
Issue Links
Activity
Now that DM-5595 has been fixed the impetus to use internal SSL is lessened (Although an inability to install openssl on the particular machine would be a problem). My main question is whether Qserv use the SSL connectivity at the moment.
Qserv itself only makes use of the hashing functions. But I don't know if mariadbclient might make use of SSL proper (e.g. talking to the mysqlproxy instance, where the qserv czar is now in-process).
I'll close this as "Won't Fix" given that the confusion from Anaconda libraries has been fixed in DM-5595 and OpenSSL is still listed as a requirement for Linux.
I don't know anything about maria's internal TLS libs so I can't really add much to this conversation. My hope is that we'll be able to link against the anaconda openssl package on OSX in the future.
Well, the only issue I see is that the system SSL libraries are updated frequently and in a timely fashion in response to CVE alerts, etc. What is the vulnerability of MariaDB's internal SSL implementation? I'd be interested in the point of view of somebody more security-conscious here (maybe Joshua Hoblitt?)
I haven't really cared too much to date about the OSX builds of qserv, since they aren't currently intended to be supported for production use. So, whatever works for the convenience of devs on OSX is okay for me, in order to get the bonus of having clang run regularly over our code.