On Thu, Jul 26, 2007 at 10:43:44AM +0200, Olaf van der Spek wrote: > Hi, > > > this is what is keeping boost 1.34.x out of testing. those long names > > (ie. libboost_program_options-gcc-mt-1_33_1.so) are the "regular" > > ones, those that come out of the build as upstream made it. it is > > explained in the Boost documentation [0]. > > Didn't Boost developers say layout=system should be used in which > -lboost_program_options works again and links with -mt?
I don't believe that is the concensus view. I think Neal Becker got it right [http://article.gmane.org/gmane.comp.lib.boost.devel/162674] I think typically linux distros want the version, but not the 'gcc41' part. We really do need to allow multiple versions, as is common practice with other shared libs on linux systems. Since multiple compilers are uncommon, distros normally ignore that. The trouble with --layout=system is that it completely strips the boost version from the library SONAME. Doing that will cause much more grief after an upgrade (because Boost doesn't maintain ABI) than the current crisis. To your second question: with --layout=system, -lboost_program_options links to the SINGLE-THREADED version. Unfortunately, this is completely opposite of previous Debian practice (for which I must take the blame). -Steve
signature.asc
Description: Digital signature
