On macOS, with both arm64: and x86_64 * the gap executable loads libgap in SageMath 10.4, 10.5 and 10.6. * crypting.so loads libgap in SageMath 10.4 and 10.6 but not 10.5.
On macOS x86_64 : * the gap executable loads libgap in SageMath 10.4, 10.5 and 10.6. * crypting.so loads libgap in SageMath 10.4 and 10.5, but not 10.6. - Marc On Wednesday, January 29, 2025 at 5:25:18 PM UTC-6 dim...@gmail.com wrote: > On Wed, Jan 29, 2025 at 4:35 PM Marc Culler <marc....@gmail.com> wrote: > > > > Looking back at the failed command I actually think this is a clang bug > which has nothing to do with GAP. The relevant option is: > > -bundle_loader /private/var/tmp/sage-10.6-current/local/bin/gap > > I think that option is broken in the latest clang. > > I think that option means that the linker should not complain about a > missing symbol if that symbol is defined in the executable which loads the > library, or in any library loaded by that executable. In the case of > local/bin/gap I see that (of course) it loads libgap: > > Why does it load libgap? That's a bug in the gap's building by Sage I > think (no idea how it happens though). > Out of the Sage tree build has the following linkage; see - no libgap > linked. > > % otool -L gap > gap: > /usr/lib/libSystem.B.dylib (compatibility version 1.0.0, > current version 1351.0.0) > /opt/homebrew/opt/gmp/lib/libgmp.10.dylib (compatibility > version 16.0.0, current version 16.0.0) > /usr/lib/libz.1.dylib (compatibility version 1.0.0, current > version 1.2.12) > /opt/homebrew/opt/readline/lib/libreadline.8.dylib > (compatibility version 8.2.0, current version 8.2.0) > > as opposed to Sage's build: > % otool -L sage/local/bin/gap > sage/local/bin/gap: > /usr/lib/libSystem.B.dylib (compatibility version 1.0.0, > current version 1351.0.0) > /opt/homebrew/opt/gmp/lib/libgmp.10.dylib (compatibility > version 16.0.0, current version 16.0.0) > /usr/lib/libz.1.dylib (compatibility version 1.0.0, current > version 1.2.12) > /opt/homebrew/opt/readline/lib/libreadline.8.dylib > (compatibility version 8.2.0, current version 8.2.0) > /Volumes/dima/sage/local/lib/libgap.9.dylib (compatibility > version 10.0.0, current version 10.0.0) > > Dima > > > LC_LOAD_DYLIB: /usr/lib/libSystem.B.dylib > > LC_LOAD_DYLIB: @rpath/libgmp.10.dylib > > LC_LOAD_DYLIB: @rpath/libz.1.dylib > > LC_LOAD_DYLIB: @rpath/libreadline.8.dylib > > LC_LOAD_DYLIB: @rpath/libgap.9.dylib > > So the linker should not have complained about those missing symbols, > much less have raised an error. And, in fact, there was no error on my > Intel system with an older clang. However, I think that linking against > libgap accomplishes the same thing, It lets the linker know that the symbol > will exist when the crypting library is loaded. So I don't think that is a > bad workaround, and I think it will not be necessary after the next release > of XCode. GAP is doing nothing wrong, except to assume that the > -bundle-loader option will work as advertised. > > > > - Marc. > > > > On Wed, Jan 29, 2025 at 3:21 PM Dima Pasechnik <dim...@gmail.com> wrote: > >> > >> > >> > >> On 29 January 2025 14:43:17 GMT-06:00, Marc Culler <marc....@gmail.com> > wrote: > >> >On Wed, Jan 29, 2025 at 1:33 PM Dima Pasechnik <dim...@gmail.com> > wrote: > >> > > >> >> > >> >> libgap is not really involved here; > >> >> > >> >cypring's GAP kernel module > >> >> (that's what's compiled here) can be loaded either in libgap, or in > >> >> gap executable - and the latter > >> >> isn't linked to libgap. > >> > > >> > > >> >I am no expert in GAP. But the code in crypting.c references symbols > which > >> >are defined in libgap. So to build crypting.so from crypting.o you > have to > >> >link against libgap or something else which defines those same symbols. > >> > >> GAP does dynamic symbol resolution when it loads an extension module. > >> On macOS, GAP does not link libgap; one can check with "otool -L". > >> > >> libgap does contain the internal symbols needed for kernel extension > modules available in GAP the executable, and is equally able to load a GAP > kernel extension module. > >> > >> If you link a GAP kernel module to libgap, and load it into GAP, the > latter will also load libgap, and the hell can potentially break loose. > >> Even if it doesn't, an extra unneeded library would be loaded. > >> > >> crypting.so, the GAP kernel extension module, normally only links to a > macOS system library, and nothing else. > >> > >> Dima > >> > >> >After building with my patch, if I run the gap executable > >> >$SAGE_ROOT/local/bin/gap then I am able to load the crypting module: > >> > > >> >% > >> > >/Applications/SageMath-10-6.app/Contents/Frameworks/Sage.framework/Versions/Current/local/bin/gap > >> > ┌───────┐ GAP 4.13.1 of 2024-06-11 > >> > │ GAP │ https://www.gap-system.org > >> > └───────┘ Architecture: aarch64-apple-darwin24-default64-kv9 > >> > Configuration: gmp 6.3.0, GASMAN, readline > >> > Loading the library and packages ... > >> > Packages: AClib 1.3.2, Alnuth 3.2.1, AtlasRep 2.1.8, AutPGrp 1.11, > >> > CRISP 1.4.6, Cryst 4.1.27, CrystCat 1.1.10, CTblLib 1.3.9, > >> > FactInt 1.6.3, FGA 1.5.0, GAPDoc 1.6.7, IO 4.8.2, > >> > IRREDSOL 1.4.4, LAGUNA 3.9.6, Polenta 1.3.10, Polycyclic 2.16, > >> > PrimGrp 3.4.4, RadiRoot 2.9, ResClasses 4.7.3, SmallGrp 1.5.3, > >> > Sophus 1.27, TomLib 1.2.11, TransGrp 3.6.5, utils 0.85 > >> > Try '??help' for help. See also '?copyright', '?cite' and '?authors' > >> >gap> LoadPackage("crypting"); > >> > >──────────────────────────────────────────────────────────────────────────── > >> >Loading crypting 0.10.4 (Hashes and Crypto in GAP) > >> >by Markus Pfeiffer (http://www.morphism.de/~markusp/). > >> >maintained by: > >> > Markus Pfeiffer (http://www.morphism.de/~markusp/) and > >> > The GAP Team (sup...@gap-system.org). > >> >Homepage: https://gap-packages.github.io/crypting/ > >> >Report issues at https://github.com/gap-packages/crypting/issues > >> > >──────────────────────────────────────────────────────────────────────────── > >> >true > >> > > >> >That is all that I know about this. > >> > > >> >- Marc > -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To view this discussion visit https://groups.google.com/d/msgid/sage-devel/bf223c11-08e5-4653-b6e5-c234c1e6a5b3n%40googlegroups.com.
