Yeah I'm glad we're now rid of the broad wildcarding mistake in symbols.map. But that was less of an issue. Functions were pulled from the right library, just with incorrect symbol versions.

On 23/03/15 19:59, Alan Griffiths wrote:
It isn't elegant to load two versions of a library but I don't think it
matters at present which version of the mir_*event_get* functions is
resolved - they simply moved (and don't depend on static data). The
alternative is to bump the client ABI (as was initially proposed in
https://code.launchpad.net/~alan-griffiths/mir/fix-1415321/+merge/253077).

FWIW We've had other silent ABI infelicities thanks to the the wildcards
in MIR_CLIENT_8 accidentally publishing symbols in the wrong stanza.

On 22/03/15 16:33, Cemil Azizoglu wrote:
Thanks Daniel for noticing this. I agree that there may be unforeseen
risks even though things seem to work. We should fix this ASAP,
preferably with a test case that screams when two versions are loaded
simultaneously. Bumping the priority of the bug so it gets proper
attention.

-C

On Fri, Mar 20, 2015 at 2:40 AM, Daniel van Vugt
<daniel.van.v...@canonical.com <mailto:daniel.van.v...@canonical.com>>
wrote:

    Maybe as a temporary solution we should have the offending symbols
    versioned as "MIR_COMMON_3.1" still from within libmirclient. At
    least till the client ABI gets bumped.

    That way existing clients are not forced to load two different
    versions of libmircommon simultaneously.



    On 20/03/15 15:32, Daniel van Vugt wrote:

        This is interesting. Since r2408 we do indeed have legacy
        clients still
        working with newer library builds. So no obvious ABI break.
        But they
        work because they can load libmircommon.so.3&4 simultaneously.

        So everything apparently still works, but I'm a little afraid
        there
        might be unseen risks with two versions of the same library in
        one process.

        [https://bugs.launchpad.net/mir/+bug/1415321/comments/3]


    --
    Mir-devel mailing list
    Mir-devel@lists.ubuntu.com <mailto:Mir-devel@lists.ubuntu.com>
    Modify settings or unsubscribe at:
    https://lists.ubuntu.com/mailman/listinfo/mir-devel




--
Cemil Azizoglu
Mir Display Server - Team Lead
Canonical USA






--
Mir-devel mailing list
Mir-devel@lists.ubuntu.com
Modify settings or unsubscribe at: 
https://lists.ubuntu.com/mailman/listinfo/mir-devel

Reply via email to