On Dec 15, 2007, at 6:28 AM, Moinak Ghosh wrote: > Alok Aggarwal wrote: >> On Fri, 14 Dec 2007, Mike Riley wrote: >> >> >> [...] >>> In the gzip_compress() function you call die() if you can't >>> locate the symbol >>> compress2() in the library, or if you can't find the library. I >>> fail to see >>> how that is different from Solaris reporting the same info and >>> not running >>> the program, unless it is that this routine is never invoked unless >>> compression is requested, in which case an uncompressed volume >>> would never >>> see this error using your new code. >>> >> >> The behavior is very different. >> >> Without this fix, lofiadm will die whether or not you >> requested compression or not with the following error >> message - >> >> ld.so.1: lofiadm: fatal: libz.so.1: open failed: No such file or >> directory >> zsh: killed lofiadm >> >> With this fix, lofiadm will error out only if compression >> is requested with the following error message - >> >> lofiadm: error locating /usr/lib/libz.so.1: No such file or directory >> >> I will let you decide which behavior you would rather see. >> > > Here is another flip question: Does the above change in the > error message worth the code change/addition being introduced ? > How much more code we will be introducing in future if/when we > decide to add support for more compression mechanisms ? > > There is precedent here. There are other utilities in SUNWcsu > that link with libraries in other packages. For eg. svccfg links > with libxml2 from SUNWlxml and libxml2 in turn links with libz. > So not having SUNWlxml and SUNWzlib will partially break smf. > There are other examples like /usr/bin/at, /usr/bin/id, > /usr/bin/pkill all in SUNWcsu. All these will break if SUNWzlib > is not installed. In this light do we need to do anything special > for lofiadm ? I think we can agree that zlib has now become critical to boot. The issue is as you point out later, how to develop a common mechanism for various compression algorithms. > > However there is yet another thing to consider. If we do introduce > another compression mechanism in future we'd be linking lofiadm > with another library. Since it is a pluggable compression > technique using dlopen seems useful, but the dlopen code should > be moved into a separate function. In addition dlopen should be > used without an absolute pathname, see dlopen(3C). > I agree.
> Also the error message should be something like: > libfoo not found, Foo compression is unavailable. > > > Regards, > Moinak. > > _______________________________________________ > caiman-discuss mailing list > caiman-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/caiman-discuss
