On Thu, Sep 3, 2015 at 12:49 AM, Ralph Castain <r...@open-mpi.org> wrote:
> George, I think you misunderstand the difference between the two modules. > PSM supports one type of fabric, and PSM2 supports a different one. They > are not interchangeable. > Ralph, what these two modules do is irrelevant. My point is that if the two MTL are build as shared libraries, and if as Gilles highlighted the dlopen support __is__ enabled then there is a simple solution that do not require a new release. > I agree with your second point. If you have a way of resolving it, I would > welcome hearing it. So far, the problems have been a lack of testing that > fails to identify problems, frequently due to missed use-cases. Perhaps > someone has a better suite of tests (or would volunteer to help develop > them), and/or would offer up access to a broader range of environments? > I wonder how many other open source software packages are getting as much testing as OMPI does? And this is without even talking about the testing marathon Paul is putting it through before each release. > As for why the distro would provide two rpm’s: you can argue that with > them. :-) > You brought the issue here to the OMPI community, I assumed you knew more about the problem. In any case it would be extremely helpful if you can pinpoint us to the discussion related to this issue where we can gather a little bit more insight. > However, I believe the issue is the same one we have with hcoll/ml and > libnl/libnl3 conflicts - and I have yet to see someone propose a reliable > fix for either of those. If you can create one, please chime in on the > libnl/libnl3 ticket as Gilles and Jeff have spent a lot of time and pain on > it. > There are few subtle differences that make the issue with PSM simpler. In the NL/NL3 case they belong to the dependence list of some of the libraries we are loading, which makes it difficult to prevent them from loading. In the case of PSM/PSM2, the MCA framework has a saying in what gets loaded (again under the assumption that we have dlopen support). George. > > > On Sep 2, 2015, at 9:38 PM, Gilles Gouaillardet <gil...@rist.or.jp> wrote: > > George, > > about your third point : > some libraries does stuff in the constructors, so "mtl = ^psm" might also > not work if OMPI was configure'd with --disable-dlopen. > as far as i know, --disable-dlopen is quite popular (and --disable-shared > --enable-static is not so much) > > Cheers, > > Gilles > > On 9/3/2015 1:31 PM, George Bosilca wrote: > > I might have missed something here but: > > 1. I bet that, and I'm certainly using a lower bound here, 99.9% of our > users will not even notice the issue between PSM and PSM2. > > 2. If there is anything that might negatively impact us as a community is > the recurrent screwed-up with our own releases. For a production-quality > software, releasing a new "stable" version every 3 weeks is not being > reactive, is being obnoxious. > > 3. Except if the distro builds OMPI statically, I see no reason to have 2 > build of OMPI due to conflicting symbols between two shared libraries that > OMPI MCA load willingly. Why a simple "mtl = ^psm" in the OMPI system wide > configuration file is not enough to solve the issue? > > George. > > > On Wed, Sep 2, 2015 at 11:47 PM, Ralph Castain <r...@open-mpi.org> wrote: > >> I’m afraid that won’t solve the problem - the distro will still feel the >> need to release -two- versions of OMPI, one with PSM and one with PSM2. >> Ordinarily, I wouldn’t care - but this creates user confusion and reflects >> on us as a community. >> >> >> > On Sep 2, 2015, at 6:50 PM, Gilles Gouaillardet <gil...@rist.or.jp> >> wrote: >> > >> > Ralph, >> > >> > what about automatically *not* building PSM2 if PSM is built and PSM2 >> is not explicitly required ? >> > /* in order to be future proof, we could even do that only if we detect >> a symbol conflict */ >> > we could abort if ompi is configure'd with both --with-psm and >> --with-psm2, or simply do nothing >> > (the end user might know what he/she is doing, and there will be >> nothing to do on the ompi side >> > when this gets fixed by the PSM folks) >> > >> > Cheers, >> > >> > Gilles >> > >> > On 9/3/2015 10:21 AM, Ralph Castain wrote: >> >> Hi folks >> >> >> >> I regret to say that 1.10.0 is hitting an issue with at least one >> upstream distro. Apparently, there is a symbol conflict between the PSM and >> PSM2 libraries that precludes building both of those MTLs at the same time. >> This is leading the distro to push for release of two OMPI 1.10.0 builds - >> one with PSM and the other with PSM2. >> >> >> >> IMO, this is a very undesirable situation. I agree with the distro >> that delaying release for some significant time as this would impact >> everyone else’s users. Therefore, assuming that the PSM team is unable to >> quickly resolve the problem in their libraries, my inclination is to >> release an immediate 1.10.1 with the PSM2 MTL removed. >> >> >> >> I’m soliciting input - any opinions? >> >> Ralph >> >> >> >> _______________________________________________ >> >> devel mailing list >> >> de...@open-mpi.org >> >> Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/devel >> >> Link to this post: >> http://www.open-mpi.org/community/lists/devel/2015/09/17919.php >> > >> > _______________________________________________ >> > devel mailing list >> > de...@open-mpi.org >> > Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/devel >> > Link to this post: >> http://www.open-mpi.org/community/lists/devel/2015/09/17920.php >> >> _______________________________________________ >> devel mailing list >> de...@open-mpi.org >> Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/devel >> Link to this post: >> http://www.open-mpi.org/community/lists/devel/2015/09/17921.php > > > > > _______________________________________________ > devel mailing listde...@open-mpi.org > Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/devel > Link to this post: > http://www.open-mpi.org/community/lists/devel/2015/09/17923.php > > > _______________________________________________ > devel mailing list > de...@open-mpi.org > Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/devel > Link to this post: > http://www.open-mpi.org/community/lists/devel/2015/09/17924.php > > > > _______________________________________________ > devel mailing list > de...@open-mpi.org > Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/devel > Link to this post: > http://www.open-mpi.org/community/lists/devel/2015/09/17925.php >
