Andreas Tille <andreas@fam-tille.de> writes: > Hi Gard, > > thanks a lot for contacting me. I'm adding the debian-r mailing list > since all our packages are team maintained and I do not want to deal > with private packages. Hi. My bad, I should have noticed that and should have contacted the list directly. > On Fri, Sep 17, 2021 at 11:15:33AM +0200, Gard Spreemann wrote: >> Hi Andreas, >> >> I randomly noticed that your src:r-cran-lbfgsb3c and my src:lbfgsb share >> significant amounts of upstream code, namely the core lbfsgb.f as >> written by Nocedal et al. >> >> I see our respective patches have diverged quite a bit. > > Currently the r-cran-lbfgsb3c package has no patches since I have not > seen any need for it. Indeed, I was mistaken – there are no Debian patches, but the CRAN source does patch the original lbfgsb, partially in ways that differ from how src:lbfgsb and SciPy upstream do it. In particular, it seems to be replacing certain state strings with numeric values. That seems reasonable, but may cause some headaches for a deduplication effort. > My only motivation to package this was that it is > used in the test of r-cran-optimx since it seems a good idea to run the > full test suite to me. > >> Do you think >> there's any hope for deduplication here? Could a wrapper conceivably be >> written that lets c-cran-lbfgsb3c use liblbfgsb0? A few years ago, the >> SciPy maintainers successfully dropped their lbfsgb.f copy in favor of >> linking with liblbfgsb0 (and in the process I adopted some of their >> patches into src:lbfgsb). > > I absolutely agree that code duplication should be avoided. There are > some bug reports against some r-cran-* packages to avoid this but I need > to admit that my workload of things with higher importance is really > high. So if you (or a reader of this list) is not able to provide a > patch to avoid the code copy I'd prefer if you file a bug report to make > sure that this issue will not be forgotten in the future. I totally understand! I will see if I can come up with something, otherwise status quo seems fine too. Best, Gard
Attachment:
signature.asc
Description: PGP signature