>>>>> Dirk Eddelbuettel writes:> On 10 May 2019 at 10:52, Johannes Ranke wrote: > | Thanks, that sounds good. But I need some help as I do not know much about > | autoconf and Debian packaging: Is it enough to patch configure.ac (r76467) or > | do we need to update configure as well (r76468)?> Again, that would happen in the sources you pick up from me, and per> part1 https://github.com/wch/r-source/commit/d60792d3f404cc970ab33ebee1d4481a770ec15c > part2 https://github.com/wch/r-source/commit/05046138c6fa9b40b9676f6b67498d74281d5030> it only affects gfortran >= 7, and my Debian stable fileserver shows gcc > still the default so no rush.> | > In principle I think all Fortran BLAS/LAPACK implementations (refblas > | > and ATLAS) packaged for buster should be recompiled with > | > -fno-optimize-sibling-calls (they may be fine in case they were compiled > | > with older version of gfortran-8, but then the next rebuild will cause > | > trouble): Dirk, any chance you could get the package maintainers to make > | > these changes? > | > | It seems to me this is of relevance for for S?bastien (Ccing), or more > | generally for debian-science.> Not really. I do not think anybody concluded our LAPACK/BLAS needed to be > recompiled. The discussion about this has been going on for a few weeks and > is now also between gcc/gfortran upstream and the lapack folks. See Tomas's > posts on (IIRC) r-devel.> So in short, things are moving, and in the right direction. Also worth > recalling that the _initial findings_ were and still are about a gcc / > gfortran version _not even in Debian unstable yet_ (but the folks in > Fedora once again jumped the gun and they now have that problem with > gfortran 9).Afaics, the issue certainly affects current gfortran-8 in testing? -k
On 10 May 2019 at 13:46, Kurt Hornik wrote: | >>>>> Dirk Eddelbuettel writes: | | > On 10 May 2019 at 10:52, Johannes Ranke wrote: | > | Thanks, that sounds good. But I need some help as I do not know much about | > | autoconf and Debian packaging: Is it enough to patch configure.ac (r76467) or | > | do we need to update configure as well (r76468)? | | > Again, that would happen in the sources you pick up from me, and per | | > part1 https://github.com/wch/r-source/commit/d60792d3f404cc970ab33ebee1d4481a770ec15c | > part2 https://github.com/wch/r-source/commit/05046138c6fa9b40b9676f6b67498d74281d5030 | | > it only affects gfortran >= 7, and my Debian stable fileserver shows gcc | > still the default so no rush. | | > | > In principle I think all Fortran BLAS/LAPACK implementations (refblas | > | > and ATLAS) packaged for buster should be recompiled with | > | > -fno-optimize-sibling-calls (they may be fine in case they were compiled | > | > with older version of gfortran-8, but then the next rebuild will cause | > | > trouble): Dirk, any chance you could get the package maintainers to make | > | > these changes? | > | | > | It seems to me this is of relevance for for S?bastien (Ccing), or more | > | generally for debian-science. | | > Not really. I do not think anybody concluded our LAPACK/BLAS needed to be | > recompiled. The discussion about this has been going on for a few weeks and | > is now also between gcc/gfortran upstream and the lapack folks. See Tomas's | > posts on (IIRC) r-devel. | | > So in short, things are moving, and in the right direction. Also worth | > recalling that the _initial findings_ were and still are about a gcc / | > gfortran version _not even in Debian unstable yet_ (but the folks in | > Fedora once again jumped the gun and they now have that problem with | > gfortran 9). | | Afaics, the issue certainly affects current gfortran-8 in testing? Correct -- my bad. As Debian pulled that upstream patch in. In any event I was planning to push an updated r-base package carrying the patch to configure. And the as-discussed issue with noisy "verboten compiler switches" warning we are being handed between Debian imposing switches and R not liking them. Dirk -- http://dirk.eddelbuettel.com | @eddelbuettel | edd at debian.org
>>>>> Dirk Eddelbuettel writes:> On 10 May 2019 at 13:46, Kurt Hornik wrote: > | >>>>> Dirk Eddelbuettel writes: > | > | > On 10 May 2019 at 10:52, Johannes Ranke wrote: > | > | Thanks, that sounds good. But I need some help as I do not know much about > | > | autoconf and Debian packaging: Is it enough to patch configure.ac (r76467) or > | > | do we need to update configure as well (r76468)? > | > | > Again, that would happen in the sources you pick up from me, and per > | > | > part1 https://github.com/wch/r-source/commit/d60792d3f404cc970ab33ebee1d4481a770ec15c > | > part2 https://github.com/wch/r-source/commit/05046138c6fa9b40b9676f6b67498d74281d5030 > | > | > it only affects gfortran >= 7, and my Debian stable fileserver shows gcc > | > still the default so no rush. > | > | > | > In principle I think all Fortran BLAS/LAPACK implementations (refblas > | > | > and ATLAS) packaged for buster should be recompiled with > | > | > -fno-optimize-sibling-calls (they may be fine in case they were compiled > | > | > with older version of gfortran-8, but then the next rebuild will cause > | > | > trouble): Dirk, any chance you could get the package maintainers to make > | > | > these changes? > | > | > | > | It seems to me this is of relevance for for S?bastien (Ccing), or more > | > | generally for debian-science. > | > | > Not really. I do not think anybody concluded our LAPACK/BLAS needed to be > | > recompiled. The discussion about this has been going on for a few weeks and > | > is now also between gcc/gfortran upstream and the lapack folks. See Tomas's > | > posts on (IIRC) r-devel. > | > | > So in short, things are moving, and in the right direction. Also worth > | > recalling that the _initial findings_ were and still are about a gcc / > | > gfortran version _not even in Debian unstable yet_ (but the folks in > | > Fedora once again jumped the gun and they now have that problem with > | > gfortran 9). > | > | Afaics, the issue certainly affects current gfortran-8 in testing?> Correct -- my bad. As Debian pulled that upstream patch in. In any > event I was planning to push an updated r-base package carrying the > patch to configure.Thanks. Btw, Tomas just told R Core that ********************************************************************* I had a quick look at reference lapack builds in different distributions: looking at the disassembly, and specifically for dposv tail-calling into dpotrs. I checked the latest packages from Fedora 30, Fedora Rawhide (the same?), Ubuntu 19.04, Debian Sid, OpenSuse Leap 42.3. None of the builds had that problem (yet). I've been looking for the packages via pkgs.org. ********************************************************************* so we should be safe re BLAS/LAPACK binary packages in testing for now. (But of course, this might change in case of rebuilds using current gfortran-8). Best -k> And the as-discussed issue with noisy "verboten compiler switches" warning we > are being handed between Debian imposing switches and R not liking them.> Dirk> -- > http://dirk.eddelbuettel.com | @eddelbuettel | edd at debian.org