ABI incompatibility

[Gianfranco Costamagna]

yes, thanks for confirming, I was already aware of this ABI change (I think I even sent a mail about this? not sure)

This is why I uploaded boinc-app-seti soonafter every architecture has been built.

Would be nice to have something like "boinc:Depends" or similar in control file, just to be sure we just need to ask ftpmaster to rebuild.

haskell team seems to do already this "expansion"

do you have time/knowledge to do it?

cheers,


Gianfranco




Il Venerdì 4 Luglio 2014 18:22, Guo Yixuan <culu.gyx at gmail.com> ha scritto:


>
>
>Hi,
>
>
>Recently I encountered a seti crash with this stderr log:
>
>
>boinc-app-seti is invoked with the following arguments: 
>--start
>--end.
>I: boinc_parse_init_data_file
>I: boinc_get_init_data
>I: sah_graphics_init
>I: boinc_init
>I: worker
>I: worker() - common_init
>I: worker() - read_wu_state
>I: worker() - seti_do_work
>Restarted at 0.41 percent.
>../../projects/setiathome.berkeley.edu/setiathome_v7: symbol lookup error: ../../projects/setiathome.berkeley.edu/setiathome_v7: undefined symbol: _Z20boinc_temporary_exitiPKc
>
>
></stderr_txt>
>
>
>The reason is I used a older seti with a newer libboinc-app7:
>
>
>$ dpkg -l libboinc-app7 boinc-app-seti-graphics 
>Desired=Unknown/Install/Remove/Purge/Hold
>| Status=Not/Inst/Conf-files/Unpacked/halF-conf/Half-inst/trig-aWait/Trig-pend
>|/ Err?=(none)/Reinst-required (Status,Err: uppercase=bad)
>||/ Name                              Version               Architecture          Description
>+++-=================================-=====================-=====================-=======================================================================
>ii  boinc-app-seti-graphics           7.28~svn2385-1        amd64                 SETI at home application for the BOINC client (with graphics)
>ii  libboinc-app7                     7.4.8+dfsg-1          amd64                 libraries for BOINC's scientific applications
>
>
>The boinc_temporary_exit() function changed its signature
>at 7.4.1 [1], which caused this ABI incompatibility. Can we
>solve this problem by using stricter version-qualified
>dependency?
>
>
>(I think the correct solution would be a SONAME change or
>symbol versioning, but these seem to be impossible without
>support from upstream.)
>
>
>[1] http://boinc.berkeley.edu/gitweb/?p=boinc-v2.git;a=commitdiff;h=f15f6d2ba0621f820bd685877b89c14b07c732f2
>
>
>
>Regards,
>
>
>Yixuan
>
>