Jeff Johnson wrote: > > On Jun 8, 2007, at 9:35 AM, Andy Green wrote: > >> Jeff Johnson wrote: >> >>> 2) absolute argv[0] path will be used if present to find >>> configuration. relative argv[0] path will be >>> made absolute using PATH lookup. >> >> That sounds like a very good idea, cf compilers finding libraries. >>
I had forgotten this until I read the email, but the other reason I went with environment variables and a shell wrapper for my relocation code was to be able to pass in LD_LIBRARY_PATH so that the correct set of the librpm libraries would be loaded. I see two ways around this, either you need to statically link the librpm libraries into whatever the users accesses, or dlopen them. (Or continue to use a shell wrapper and pickup the LD_LIBRARY_PATH.) --Mark > Nod. I ook to gcc for ideas and prior art always. E.g. rpm macros, > particularly the primitive > test syntax like %{?...}, are ideas that were used from gcc option > passing to > back-end helpers long ago. The implementation has nothing to do with gcc, > blame me, not gcc for that. > > Meanwhile, one important difference is that not only a rpm executable, > but also rpmlib bindings, need configuration "discovery". That may mean > I have to map sonames -> paths (using ldconfig -p) for finding $RPMHOME. > > That is likely be more controversial than finding executable in PATH and > using a relative path from that location. > > 73 de Jeff > ______________________________________________________________________ > RPM Package Manager http://rpm5.org > Developer Communication List rpm-devel@rpm5.org ______________________________________________________________________ RPM Package Manager http://rpm5.org Developer Communication List rpm-devel@rpm5.org