I think I will ping pynac upstream to see if they have any idea but I think it may be a recurring gcc or a subtle abuse of C++. It would be interesting to see if putting -fno-fat-lto-objects would with getting the symbol in pynac. There are other things ‘i can think off that may work but need to consult upstream and people who know more c++.
François > On 26/11/2014, at 18:57, Jeramia Poland <[email protected]> wrote: > > The output of > > nm -D /usr/lib64/python2.7/site-packages/sage/symbolic/pynac.so | grep > ZTVN5GiNaC9containerISt6vectorEE > > is: > > U _ZTVN5GiNaC9containerISt6vectorEE (with the ZTVN part showing up in red) > > The output of > > nm -D /usr/lib64/libpynac.so.1.1.0 | grep ZTVN5GiNaC9containerISt6vectorEE > > is: > > (nothing it appears to be missing) > > Output of the third command is attached. The -ffat-lto-objects flag should > be in there. I added after the first couple of packages (like pciutils, > elfutils, openssh, etc.) wouldn't build to see if it would help so I didn't > have to use the "no-lto" environment setting from that graphite and lto > instruction page I referenced earlier. It didn't see to work as I now have a > package.env file full of no-lto entries. Neither sage nor pynac are no-lto > entries right now as it appeared to me that they merged. > > Jeramia > > > > On Wed, Nov 26, 2014 at 2:43 AM, François Bissey <[email protected]> wrote: > I think I have all the logs I can use at the moment. I want the output of the > following two commands: > nm -D /usr/lib64/python2.7/site-packages/sage/symbolic/pynac.so | grep > ZTVN5GiNaC9containerISt6vectorEE > > nm -D /usr/lib64/libpynac.so.1.1.0 | grep ZTVN5GiNaC9containerISt6vectorEE > > Actually make that 3 commands with the output of > readelf -d /usr/lib64/python2.7/site-packages/sage/symbolic/pynac.so > > Also to cross check that I am not missing anything fro your info output, > I cannot see "-ffat-lto-objects" anywhere in your compilation flags. > Is this correct? > > Francois > > On Mon, 24 Nov 2014 19:05:42 François Bissey wrote: > > Hum I am starting to wonder if there are things going on between lot and C++ > > templates. More professional code may do ok but a lot of the stuff we have > > in sage may be a bit dubious. Can you send me the penal build log please > > and for good measure the one for sage. You can do it privately if you don’t > > want to post big file on the list. > > > > I have been free of grading for almost 4 years. My wife is very happy about > > that. > > > > François > > > > > On 24/11/2014, at 17:12, Jeramia Poland <[email protected]> wrote: > > > > > > Sorry for the delayed reply, I had a bunch of papers to grade. Yes, it is > > > installed. Attached are the lld results. Jeramia > > > > > > On Sun, Nov 23, 2014 at 1:20 AM, François Bissey <[email protected]> > > > wrote: I hate automated spellchecking, I meant pynac not “penal” I > > > remember turning it off on an iPad after a ludicrous suggestion for > > > lapack. > > > > > > François > > > > > > > On 22/11/2014, at 15:33, François Bissey <[email protected]> wrote: > > > > I assume penal is installed otherwise > > > > > > <lib_link_1.txt><lib_link_2.txt> > > > > <readelf.txt>
