On Wed, Feb 26, 2025 at 11:27 AM Marc Culler <[email protected]> wrote:
>
> What about just checking the return status of nm piped to grep, as we were 
> just doing?  (Once you know which function you need to look for).

the function name is correct, the mangled name is dgeqrf_, as reported
by AC_FC_FUNC,
with this value, AC_CHECK_FUNC passes.

However, I found something that looks like an upstream (Homebrew?)
error: "-fopenmp" is pushed into CFLAGS
from the pkg-config, but this causes C compilations (with clang) fail,
as Apple clang does not support openmp,
as far as I can tell (and Sage's tests on gcc agree with me here).

It's probably due to Homebrew using a compiler which can ignore such
an unknown option.

% pkg-config -cflags openblas
-I/opt/homebrew/Cellar/openblas/0.3.29/include -fopenmp

Note that without OpenMP openblas is much, much less performant than with it.
According to the great R project: https://mac.r-project.org/openmp/,
on XCode clang openmp is supported with
-Xclang -fopenmp
and one also needs an openmp dylib - which Apple doesn't ship (thanks
Apple, as usual - of course they want us to use their propriatory
stuff),
but it is available from R project at the URL above, or from
https://formulae.brew.sh/formula/libomp

I have yet to try the latter. Probably the flags in openblas.pc still
need to be corrected.

Dima

>
> - Marc
>
> On Wed, Feb 26, 2025 at 11:14 AM Dima Pasechnik <[email protected]> wrote:
>>
>> On Wed, Feb 26, 2025 at 10:44 AM Marc Culler <[email protected]> wrote:
>> >
>> > The code that is rejecting openblas 0.3.29 appears to be in 
>> > build/pkgs/openblas/spkgconfigure.m4.
>> >
>> > That file includes tests for the existence of the symbol dgeqrf_ for the 
>> > purpose of deciding whether a version of openblas also provides lapack.  
>> > According to github blame, that seems to have been Dima's idea 5 or 6 
>> > years ago.  Given that LAPACK appears so many times when searching for 
>> > dgeqfr in the output of nm, I would guess that version 0.3.29 does in fact 
>> > provide lapack.  But the tests in that m4 file need to be revised.
>>
>> yes, right, that's exactly what I'm doing now. However, the test is
>> very tricky and complicated, as
>> a standard way to check that a Fortran function is available is to call 
>> Fortran
>> (something what autoconf macro AC_FC_FUNC would do)
>> Instead, the test is done with C++, is much more complicated, and
>> fails on macOS arm64 for reasons
>> not yet known to me.
>>
>> I'm trying to see if at least AC_FC_FUNC can do the job, in the
>> presence of gfortran...
>>
>> Dima
>>
>> >
>> > - Marc
>> >
>> > On Wednesday, February 26, 2025 at 7:45:36 AM UTC-6 [email protected] 
>> > wrote:
>> >>
>> >> indeed, I get
>> >>
>> >> MAC-04017247:sage dcoudert$ nm 
>> >> /opt/homebrew/Cellar/openblas/0.3.29/lib/libopenblas.dylib | grep  dgeqrfp
>> >>
>> >> 0000000000d84dbc T _LAPACKE_dgeqrfp
>> >>
>> >> 0000000000d84eec T _LAPACKE_dgeqrfp_work
>> >>
>> >> 00000000008f4120 T _dgeqrfp_
>> >>
>> >>
>> >>
>> >> On Tuesday, February 25, 2025 at 11:11:21 PM UTC+1 [email protected] 
>> >> wrote:
>> >>>
>> >>> The symbol which is reported as missing from openblas 0.3.29 (dgeqrf_) 
>> >>> is defined in openblas 0.3.28 (which is the version built by the Sage 
>> >>> openblas spkg).
>> >>> % nm local/lib/libopenblas.dylib | grep dgeqrf
>> >>> 00000000007fcb64 T _LAPACKE_dgeqrf
>> >>> 00000000007fcca0 T _LAPACKE_dgeqrf_work
>> >>> 00000000007fce58 T _LAPACKE_dgeqrfp
>> >>> 00000000007fcf94 T _LAPACKE_dgeqrfp_work
>> >>> 0000000000364c60 T _dgeqrf_
>> >>> 00000000003651a0 T _dgeqrfp_
>> >>>
>> >>> If you run that nm command on the Homebrew version of libopenblas.dylib 
>> >>> it will probably not show that symbol.
>> >>>
>> >>> The configure script author seemed to think that dgeqrf_ is a good 
>> >>> symbol to check for when deciding whether an openblas library is 
>> >>> acceptable.  But that symbol was apparently removed from 0.3.29.  So 
>> >>> maybe that was just a bad choice and the script should be checking for 
>> >>> something else instead.
>> >>>
>> >>> - Marc
>> >>>
>> >>> On Tuesday, February 25, 2025 at 3:22:00 PM UTC-6 [email protected] 
>> >>> wrote:
>> >>>>
>> >>>>
>> >>>> MAC-04017247:sage dcoudert$ otool -TV 
>> >>>> /opt/homebrew/Cellar/openblas/0.3.29/lib/libopenblas.dylib
>> >>>>
>> >>>> /opt/homebrew/Cellar/openblas/0.3.29/lib/libopenblas.dylib:
>> >>>>
>> >>>> Table of contents (0 entries)
>> >>>>
>> >>>> module name      symbol name
>> >>>>
>> >>>> MAC-04017247:sage dcoudert$ otool -TV 
>> >>>> /opt/homebrew/Cellar/openblas/0.3.29/lib/libopenblas.a | grep dgeqrf
>> >>>>
>> >>>> /opt/homebrew/Cellar/openblas/0.3.29/lib/libopenblas.a(dgeqrf.o):
>> >>>>
>> >>>> /opt/homebrew/Cellar/openblas/0.3.29/lib/libopenblas.a(dgeqrfp.o):
>> >>>>
>> >>>> /opt/homebrew/Cellar/openblas/0.3.29/lib/libopenblas.a(lapacke_dgeqrf.o):
>> >>>>
>> >>>> /opt/homebrew/Cellar/openblas/0.3.29/lib/libopenblas.a(lapacke_dgeqrf_work.o):
>> >>>>
>> >>>> /opt/homebrew/Cellar/openblas/0.3.29/lib/libopenblas.a(lapacke_dgeqrfp.o):
>> >>>>
>> >>>> /opt/homebrew/Cellar/openblas/0.3.29/lib/libopenblas.a(lapacke_dgeqrfp_work.o):
>> >>>>
>> >>>>
>> >>>> and yes, I'm trying with https://github.com/sagemath/sage/pull/39571
>> >>>>
>> >>>> On Tuesday, February 25, 2025 at 8:59:05 PM UTC+1 Dima Pasechnik wrote:
>> >>>>>
>> >>>>> On Tue, Feb 25, 2025 at 1:01 PM [email protected]
>> >>>>> <[email protected]> wrote:
>> >>>>> >
>> >>>>> > my homebrew install of openblas (0.3.29) is not accepted, as can be 
>> >>>>> > seen in the config.log file.
>> >>>>>
>> >>>>> can you find anything like "dgeqrf" in the output of
>> >>>>>
>> >>>>> otool -TV /opt/homebrew/Cellar/openblas/0.3.29/lib/libopenblas.dylib
>> >>>>>
>> >>>>> not 100% sure about the exact name of the dylib , but you get the 
>> >>>>> idea, right?
>> >>>>> The issue comes from
>> >>>>> https://github.com/sagemath/sagetrac-mirror/commit/71ac6a5e18ad2e08383a122ef305df593a34aa3a
>> >>>>> which is a very hacky way to avoid the use of gfortran. We can re-work
>> >>>>> it if needed.
>> >>>>>
>> >>>>> Dima
>> >>>>> >
>> >>>>> > ## --------------------------------------------------------- ##
>> >>>>> > ## Checking whether SageMath should install SPKG openblas... ##
>> >>>>> > ## --------------------------------------------------------- ##
>> >>>>> > configure:18884: checking whether any of gfortran is installed as or 
>> >>>>> > will be installed as SPKG
>> >>>>> > configure:18894: result: no
>> >>>>> > configure:18908: checking for openblas >= 0.2.20 openblas < 0.3.99
>> >>>>> > configure:18915: $PKG_CONFIG --exists --print-errors "openblas >= 
>> >>>>> > 0.2.20 openblas < 0.3.99"
>> >>>>> > configure:18918: $? = 0
>> >>>>> > configure:18932: $PKG_CONFIG --exists --print-errors "openblas >= 
>> >>>>> > 0.2.20 openblas < 0.3.99"
>> >>>>> > configure:18935: $? = 0
>> >>>>> > configure:19931: result: yes
>> >>>>> > configure:19940: $PKG_CONFIG --exists --print-errors "openblas"
>> >>>>> > configure:19943: $? = 0
>> >>>>> > configure:19963: checking for cblas_dgemm
>> >>>>> > configure:19963: g++ -std=gnu++11 -std=gnu++11 -o conftest 
>> >>>>> > conftest.cpp -L/opt/homebrew/Cellar/openblas/0.3.29/lib -lopenblas 
>> >>>>> > -lbz2 -lglpk -lgmp -lm >&5
>> >>>>> > configure:19963: $? = 0
>> >>>>> > configure:19963: result: yes
>> >>>>> > configure:20044: checking for dgeqrf
>> >>>>> > configure:20044: g++ -std=gnu++11 -std=gnu++11 -o conftest 
>> >>>>> > conftest.cpp -L/opt/homebrew/Cellar/openblas/0.3.29/lib -lopenblas 
>> >>>>> > -lbz2 -lglpk -lgmp -lm >&5
>> >>>>> > Undefined symbols for architecture arm64:
>> >>>>> > "_dgeqrf", referenced from:
>> >>>>> > _main in conftest-e6c8b5.o
>> >>>>> > ld: symbol(s) not found for architecture arm64
>> >>>>> > clang++: error: linker command failed with exit code 1 (use -v to 
>> >>>>> > see invocation)
>> >>>>> > configure:20044: $? = 1
>> >>>>> > configure: failed program was:
>> >>>>> > | /* confdefs.h */
>> >>>>> > | #define PACKAGE_NAME "Sage"
>> >>>>> > | #define PACKAGE_TARNAME "sage"
>> >>>>> > | #define PACKAGE_VERSION "10.6.beta7"
>> >>>>> > | #define PACKAGE_STRING "Sage 10.6.beta7"
>> >>>>> > | #define PACKAGE_BUGREPORT "[email protected]"
>> >>>>> > | #define PACKAGE_URL ""
>> >>>>> > | #define PACKAGE "sage"
>> >>>>> > | #define VERSION "10.6.beta7"
>> >>>>> > | #define HAVE_STDIO_H 1
>> >>>>> > | #define HAVE_STDLIB_H 1
>> >>>>> > | #define HAVE_STRING_H 1
>> >>>>> > | #define HAVE_INTTYPES_H 1
>> >>>>> > | #define HAVE_STDINT_H 1
>> >>>>> > | #define HAVE_STRINGS_H 1
>> >>>>> > | #define HAVE_SYS_STAT_H 1
>> >>>>> > | #define HAVE_SYS_TYPES_H 1
>> >>>>> > | #define HAVE_UNISTD_H 1
>> >>>>> > | #define STDC_HEADERS 1
>> >>>>> > | #define HAVE_LIBM 1
>> >>>>> > | #define HAVE_CXX11 1
>> >>>>> > | #define HAVE_GMP_H 1
>> >>>>> > | #define ABSOLUTE_GMP_H "///opt/homebrew/include/gmp.h"
>> >>>>> > | #define HAVE_BOOST /**/
>> >>>>> > | /* end confdefs.h. */
>> >>>>> > | /* Define dgeqrf to an innocuous variant, in case <limits.h> 
>> >>>>> > declares dgeqrf.
>> >>>>> > | For example, HP-UX 11i <limits.h> declares gettimeofday. */
>> >>>>> > | #define dgeqrf innocuous_dgeqrf
>> >>>>> > |
>> >>>>> > | /* System header to define __stub macros and hopefully few 
>> >>>>> > prototypes,
>> >>>>> > | which can conflict with char dgeqrf (void); below. */
>> >>>>> > |
>> >>>>> > | #include <limits.h>
>> >>>>> > | #undef dgeqrf
>> >>>>> > |
>> >>>>> > | /* Override any GCC internal prototype to avoid an error.
>> >>>>> > | Use char because int might match the return type of a GCC
>> >>>>> > | builtin and then its argument prototype would still apply. */
>> >>>>> > | #ifdef __cplusplus
>> >>>>> > | extern "C"
>> >>>>> > | #endif
>> >>>>> > | char dgeqrf (void);
>> >>>>> > | /* The GNU C library defines this for functions which it implements
>> >>>>> > | to always fail with ENOSYS. Some functions are actually named
>> >>>>> > | something starting with __ and the normal name is an alias. */
>> >>>>> > | #if defined __stub_dgeqrf || defined __stub___dgeqrf
>> >>>>> > | choke me
>> >>>>> > | #endif
>> >>>>> > |
>> >>>>> > | int
>> >>>>> > | main (void)
>> >>>>> > | {
>> >>>>> > | return dgeqrf ();
>> >>>>> > | ;
>> >>>>> > | return 0;
>> >>>>> > | }
>> >>>>> > configure:20044: result: no
>> >>>>> > configure:20051: checking for dgeqrf_
>> >>>>> > configure:20051: g++ -std=gnu++11 -std=gnu++11 -o conftest 
>> >>>>> > conftest.cpp -L/opt/homebrew/Cellar/openblas/0.3.29/lib -lopenblas 
>> >>>>> > -lbz2 -lglpk -lgmp -lm >&5
>> >>>>> > configure:20051: $? = 0
>> >>>>> > configure:20051: result: yes
>> >>>>> > configure:20058: checking for DGEQRF
>> >>>>> > configure:20058: g++ -std=gnu++11 -std=gnu++11 -o conftest 
>> >>>>> > conftest.cpp -L/opt/homebrew/Cellar/openblas/0.3.29/lib -lopenblas 
>> >>>>> > -lbz2 -lglpk -lgmp -lm >&5
>> >>>>> > Undefined symbols for architecture arm64:
>> >>>>> > "_DGEQRF", referenced from:
>> >>>>> > _main in conftest-8dfd93.o
>> >>>>> > ld: symbol(s) not found for architecture arm64
>> >>>>> > clang++: error: linker command failed with exit code 1 (use -v to 
>> >>>>> > see invocation)
>> >>>>> > configure:20058: $? = 1
>> >>>>> > configure: failed program was:
>> >>>>> > | /* confdefs.h */
>> >>>>> > | #define PACKAGE_NAME "Sage"
>> >>>>> > | #define PACKAGE_TARNAME "sage"
>> >>>>> > | #define PACKAGE_VERSION "10.6.beta7"
>> >>>>> > | #define PACKAGE_STRING "Sage 10.6.beta7"
>> >>>>> > | #define PACKAGE_BUGREPORT "[email protected]"
>> >>>>> > | #define PACKAGE_URL ""
>> >>>>> > | #define PACKAGE "sage"
>> >>>>> > | #define VERSION "10.6.beta7"
>> >>>>> > | #define HAVE_STDIO_H 1
>> >>>>> > | #define HAVE_STDLIB_H 1
>> >>>>> > | #define HAVE_STRING_H 1
>> >>>>> > | #define HAVE_INTTYPES_H 1
>> >>>>> > | #define HAVE_STDINT_H 1
>> >>>>> > | #define HAVE_STRINGS_H 1
>> >>>>> > | #define HAVE_SYS_STAT_H 1
>> >>>>> > | #define HAVE_SYS_TYPES_H 1
>> >>>>> > | #define HAVE_UNISTD_H 1
>> >>>>> > | #define STDC_HEADERS 1
>> >>>>> > | #define HAVE_LIBM 1
>> >>>>> > | #define HAVE_CXX11 1
>> >>>>> > | #define HAVE_GMP_H 1
>> >>>>> > | #define ABSOLUTE_GMP_H "///opt/homebrew/include/gmp.h"
>> >>>>> > | #define HAVE_BOOST /**/
>> >>>>> > | /* end confdefs.h. */
>> >>>>> > | /* Define DGEQRF to an innocuous variant, in case <limits.h> 
>> >>>>> > declares DGEQRF.
>> >>>>> > | For example, HP-UX 11i <limits.h> declares gettimeofday. */
>> >>>>> > | #define DGEQRF innocuous_DGEQRF
>> >>>>> > |
>> >>>>> > | /* System header to define __stub macros and hopefully few 
>> >>>>> > prototypes,
>> >>>>> > | which can conflict with char DGEQRF (void); below. */
>> >>>>> > |
>> >>>>> > | #include <limits.h>
>> >>>>> > | #undef DGEQRF
>> >>>>> > |
>> >>>>> > | /* Override any GCC internal prototype to avoid an error.
>> >>>>> > | Use char because int might match the return type of a GCC
>> >>>>> > | builtin and then its argument prototype would still apply. */
>> >>>>> > | #ifdef __cplusplus
>> >>>>> > | extern "C"
>> >>>>> > | #endif
>> >>>>> > | char DGEQRF (void);
>> >>>>> > | /* The GNU C library defines this for functions which it implements
>> >>>>> > | to always fail with ENOSYS. Some functions are actually named
>> >>>>> > | something starting with __ and the normal name is an alias. */
>> >>>>> > | #if defined __stub_DGEQRF || defined __stub___DGEQRF
>> >>>>> > | choke me
>> >>>>> > | #endif
>> >>>>> > |
>> >>>>> > | int
>> >>>>> > | main (void)
>> >>>>> > | {
>> >>>>> > | return DGEQRF ();
>> >>>>> > | ;
>> >>>>> > | return 0;
>> >>>>> > | }
>> >>>>> > configure:20058: result: no
>> >>>>> > configure:20065: checking for DGEQRF_
>> >>>>> > configure:20065: g++ -std=gnu++11 -std=gnu++11 -o conftest 
>> >>>>> > conftest.cpp -L/opt/homebrew/Cellar/openblas/0.3.29/lib -lopenblas 
>> >>>>> > -lbz2 -lglpk -lgmp -lm >&5
>> >>>>> > Undefined symbols for architecture arm64:
>> >>>>> > "_DGEQRF_", referenced from:
>> >>>>> > _main in conftest-ff1fe7.o
>> >>>>> > ld: symbol(s) not found for architecture arm64
>> >>>>> > clang++: error: linker command failed with exit code 1 (use -v to 
>> >>>>> > see invocation)
>> >>>>> > configure:20065: $? = 1
>> >>>>> > configure: failed program was:
>> >>>>> > | /* confdefs.h */
>> >>>>> > | #define PACKAGE_NAME "Sage"
>> >>>>> > | #define PACKAGE_TARNAME "sage"
>> >>>>> > | #define PACKAGE_VERSION "10.6.beta7"
>> >>>>> > | #define PACKAGE_STRING "Sage 10.6.beta7"
>> >>>>> > | #define PACKAGE_BUGREPORT "[email protected]"
>> >>>>> > | #define PACKAGE_URL ""
>> >>>>> > | #define PACKAGE "sage"
>> >>>>> > | #define VERSION "10.6.beta7"
>> >>>>> > | #define HAVE_STDIO_H 1
>> >>>>> > | #define HAVE_STDLIB_H 1
>> >>>>> > | #define HAVE_STRING_H 1
>> >>>>> > | #define HAVE_INTTYPES_H 1
>> >>>>> > | #define HAVE_STDINT_H 1
>> >>>>> > | #define HAVE_STRINGS_H 1
>> >>>>> > | #define HAVE_SYS_STAT_H 1
>> >>>>> > | #define HAVE_SYS_TYPES_H 1
>> >>>>> > | #define HAVE_UNISTD_H 1
>> >>>>> > | #define STDC_HEADERS 1
>> >>>>> > | #define HAVE_LIBM 1
>> >>>>> > | #define HAVE_CXX11 1
>> >>>>> > | #define HAVE_GMP_H 1
>> >>>>> > | #define ABSOLUTE_GMP_H "///opt/homebrew/include/gmp.h"
>> >>>>> > | #define HAVE_BOOST /**/
>> >>>>> > | /* end confdefs.h. */
>> >>>>> > | /* Define DGEQRF_ to an innocuous variant, in case <limits.h> 
>> >>>>> > declares DGEQRF_.
>> >>>>> > | For example, HP-UX 11i <limits.h> declares gettimeofday. */
>> >>>>> > | #define DGEQRF_ innocuous_DGEQRF_
>> >>>>> > |
>> >>>>> > | /* System header to define __stub macros and hopefully few 
>> >>>>> > prototypes,
>> >>>>> > | which can conflict with char DGEQRF_ (void); below. */
>> >>>>> > |
>> >>>>> > | #include <limits.h>
>> >>>>> > | #undef DGEQRF_
>> >>>>> > |
>> >>>>> > | /* Override any GCC internal prototype to avoid an error.
>> >>>>> > | Use char because int might match the return type of a GCC
>> >>>>> > | builtin and then its argument prototype would still apply. */
>> >>>>> > | #ifdef __cplusplus
>> >>>>> > | extern "C"
>> >>>>> > | #endif
>> >>>>> > | char DGEQRF_ (void);
>> >>>>> > | /* The GNU C library defines this for functions which it implements
>> >>>>> > | to always fail with ENOSYS. Some functions are actually named
>> >>>>> > | something starting with __ and the normal name is an alias. */
>> >>>>> > | #if defined __stub_DGEQRF_ || defined __stub___DGEQRF_
>> >>>>> > | choke me
>> >>>>> > | #endif
>> >>>>> > |
>> >>>>> > | int
>> >>>>> > | main (void)
>> >>>>> > | {
>> >>>>> > | return DGEQRF_ ();
>> >>>>> > | ;
>> >>>>> > | return 0;
>> >>>>> > | }
>> >>>>> > configure:20065: result: no
>> >>>>> > configure:20157: checking the OpenBLAS version using 
>> >>>>> > openblas_get_config
>> >>>>> > configure:20185: gcc -o conftest 
>> >>>>> > -I/opt/homebrew/Cellar/openblas/0.3.29/include -fopenmp -g -O2 
>> >>>>> > conftest.c -L/opt/homebrew/Cellar/openblas/0.3.29/lib -lopenblas 
>> >>>>> > -lbz2 -lglpk -lgmp -lm >&5
>> >>>>> > clang: error: unsupported option '-fopenmp'
>> >>>>> > clang: error: unsupported option '-fopenmp'
>> >>>>> > configure:20185: $? = 1
>> >>>>> > configure: program exited with status 1
>> >>>>> > configure: failed program was:
>> >>>>> > | /* confdefs.h */
>> >>>>> > | #define PACKAGE_NAME "Sage"
>> >>>>> > | #define PACKAGE_TARNAME "sage"
>> >>>>> > | #define PACKAGE_VERSION "10.6.beta7"
>> >>>>> > | #define PACKAGE_STRING "Sage 10.6.beta7"
>> >>>>> > | #define PACKAGE_BUGREPORT "[email protected]"
>> >>>>> > | #define PACKAGE_URL ""
>> >>>>> > | #define PACKAGE "sage"
>> >>>>> > | #define VERSION "10.6.beta7"
>> >>>>> > | #define HAVE_STDIO_H 1
>> >>>>> > | #define HAVE_STDLIB_H 1
>> >>>>> > | #define HAVE_STRING_H 1
>> >>>>> > | #define HAVE_INTTYPES_H 1
>> >>>>> > | #define HAVE_STDINT_H 1
>> >>>>> > | #define HAVE_STRINGS_H 1
>> >>>>> > | #define HAVE_SYS_STAT_H 1
>> >>>>> > | #define HAVE_SYS_TYPES_H 1
>> >>>>> > | #define HAVE_UNISTD_H 1
>> >>>>> > | #define STDC_HEADERS 1
>> >>>>> > | #define HAVE_LIBM 1
>> >>>>> > | #define HAVE_CXX11 1
>> >>>>> > | #define HAVE_GMP_H 1
>> >>>>> > | #define ABSOLUTE_GMP_H "///opt/homebrew/include/gmp.h"
>> >>>>> > | #define HAVE_BOOST /**/
>> >>>>> > | /* end confdefs.h. */
>> >>>>> > | #include <string.h>
>> >>>>> > | char *openblas_get_config(void);
>> >>>>> > | int
>> >>>>> > | main (void)
>> >>>>> > | {
>> >>>>> > | if (!strncmp(openblas_get_config(), "OpenBLAS 0.3.22", 15)) return 
>> >>>>> > 1;
>> >>>>> > | ;
>> >>>>> > | return 0;
>> >>>>> > | }
>> >>>>> > |
>> >>>>> > configure:20191: result: known bad version
>> >>>>> > configure:20261: no suitable system package found for SPKG openblas
>> >>>>> >
>> >>>>> >
>> >>>>> >
>> >>>>> >
>> >>>>> > On Tuesday, February 25, 2025 at 7:36:27 PM UTC+1 Dima Pasechnik 
>> >>>>> > wrote:
>> >>>>> >>
>> >>>>> >> On Tue, Feb 25, 2025 at 12:03 PM [email protected]
>> >>>>> >> <[email protected]> wrote:
>> >>>>> >> >
>> >>>>> >> > I agree that we should use as many homebrew packages as possible, 
>> >>>>> >> > but how can we ensure that we follow the upgrades ?
>> >>>>> >> > For instance the versions of openblas (0.3.29), gsl (2.8), numpy 
>> >>>>> >> > (2.2.3), scipy (1.15.2) and pari (2.17.1) are newer on homebrew 
>> >>>>> >> > than what we accept.
>> >>>>> >>
>> >>>>> >> Because the update of pari is stalled due to a bug reported?
>> >>>>> >> A Pari expert should step up and tell whether it's an upstream bug, 
>> >>>>> >> or not.
>> >>>>> >> https://github.com/sagemath/sage/pull/38749
>> >>>>> >>
>> >>>>> >> openblas (of any version less than 0.3.99) should be accepted,
>> >>>>> >> provided it is recognised by pkg-config
>> >>>>> >> (you'd need to run source .homebrew-build-env for the latter).
>> >>>>> >> Same applies to gsl.
>> >>>>> >>
>> >>>>> >> numpy and scipy are Python packages, you need
>> >>>>> >>
>> >>>>> >> ./configure --enable-system-site-packages
>> >>>>> >>
>> >>>>> >> to have them from "outside".
>> >>>>> >> There are no version constraints on them imposed.
>> >>>>> >> I have a meanwhile bitrotten PR to upgrade numpy and scipy, and
>> >>>>> >> despite requesting review from 5 people, only two responded.
>> >>>>> >> https://github.com/sagemath/sage/pull/38846, back in Oct 2024.
>> >>>>> >> Then reviewers just disappeared there...
>> >>>>> >> Some of these updates are now in separate PRs. Some need to be done.
>> >>>>> >>
>> >>>>> >> See, that's what I mean by "Sage has too many vendored packages".
>> >>>>> >> These packages are bog-standard PyPI packages, we could have 
>> >>>>> >> followed
>> >>>>> >> the new versions if we didn't have this so beloved by some version
>> >>>>> >> locks.
>> >>>>> >>
>> >>>>> >> Dima
>> >>>>> >>
>> >>>>> >>
>> >>>>> >>
>> >>>>> >>
>> >>>>> >>
>> >>>>> >> >
>> >>>>> >> > I'm currently facing the following issue:
>> >>>>> >> > ipykernel fails to compile and reports that it is unable to find 
>> >>>>> >> > jupyter_core. But configure says:
>> >>>>> >> > jupyter_core: using system package; SPKG will not be installed
>> >>>>> >> >
>> >>>>> >> > I'm currently trying to compile from scratch with 
>> >>>>> >> > --with-system-jupyter_core=no. We'll see if it's better.
>> >>>>> >> >
>> >>>>> >> > On Tuesday, February 25, 2025 at 6:16:24 PM UTC+1 John H Palmieri 
>> >>>>> >> > wrote:
>> >>>>> >> >>
>> >>>>> >> >> Maybe this is naive, but if you're testing a change from the old 
>> >>>>> >> >> situation, where Python 3.13 was not allowed, to now allowing 
>> >>>>> >> >> it, I might imagine that you would test it on standard supported 
>> >>>>> >> >> platforms, like OS X with Python 3.13 installed via homebrew. We 
>> >>>>> >> >> should back out the change until it's implemented correctly.
>> >>>>> >> >>
>> >>>>> >> >>
>> >>>>> >> >> On Tuesday, February 25, 2025 at 6:46:41 AM UTC-8 Dima Pasechnik 
>> >>>>> >> >> wrote:
>> >>>>> >> >>>
>> >>>>> >> >>>
>> >>>>> >> >>>
>> >>>>> >> >>> On 24 February 2025 23:52:36 GMT-06:00, John H Palmieri 
>> >>>>> >> >>> <[email protected]> wrote:
>> >>>>> >> >>> >Sage should work out of the box without 
>> >>>>> >> >>> >"--enable-system-site-packages".
>> >>>>> >> >>>
>> >>>>> >> >>>
>> >>>>> >> >>> The box has become so heavy that one needs a forklift to do 
>> >>>>> >> >>> anything with it :-)
>> >>>>> >> >>>
>> >>>>> >> >>>
>> >>>>> >> >>> > If
>> >>>>> >> >>> >it doesn't, that's a bug in Sage. How about we don't allow 
>> >>>>> >> >>> >Python 3.13 if
>> >>>>> >> >>> >it's broken? Why wasn't this caught in testing?
>> >>>>> >> >>>
>> >>>>> >> >>> well, Sage with Python 3.13 works on Linux,
>> >>>>> >> >>> and is default on some.
>> >>>>> >> >>>
>> >>>>> >> >>> It appears that there are dozens of macOS/apple toolchain/apple 
>> >>>>> >> >>> hardware combos, and each of them has its very own set of bugs, 
>> >>>>> >> >>> misfeatures, multiple dodgy Python installs, etc.
>> >>>>> >> >>>
>> >>>>> >> >>> We can't possibly test for all of them.
>> >>>>> >> >>>
>> >>>>> >> >>> And please don't force on Sage developers the need to dig up 
>> >>>>> >> >>> and reproduce solutions by Homebrew, which has 10 times more 
>> >>>>> >> >>> developers and nontrivial funding, solutions which are already 
>> >>>>> >> >>> there.
>> >>>>> >> >>>
>> >>>>> >> >>> What's wrong with using the flag above, if it works?
>> >>>>> >> >>>
>> >>>>> >> >>> I hope William S. forgives me if I quote his up to date opinion 
>> >>>>> >> >>> on Macs:
>> >>>>> >> >>>
>> >>>>> >> >>> -----
>> >>>>> >> >>> For me, Macs are reasonable as basically "Chromebooks" that can 
>> >>>>> >> >>> also
>> >>>>> >> >>> run Linux VM's pretty well and keep track of photos and video. 
>> >>>>> >> >>> Beyond
>> >>>>> >> >>> that it's extremely painful, and I vastly prefer Linux for all 
>> >>>>> >> >>> things
>> >>>>> >> >>> dev/server/remote.
>> >>>>> >> >>> -----
>> >>>>> >> >>>
>> >>>>> >> >>> My opinion is similar - except that the only use of Mac I have 
>> >>>>> >> >>> is to cause myself extreme pain by trying to fix never ending 
>> >>>>> >> >>> macOS - specific
>> >>>>> >> >>> issues.
>> >>>>> >> >>>
>> >>>>> >> >>> On macOS, either we embrace Homebrew and use it to get external 
>> >>>>> >> >>> dependencies, or we use Conda. Current "build everything from 
>> >>>>> >> >>> ground up" has become next to impossible in any reliable way.
>> >>>>> >> >>>
>> >>>>> >> >>>
>> >>>>> >> >>> Dima
>> >>>>> >> >>>
>> >>>>> >> >>>
>> >>>>> >> >>> >
>> >>>>> >> >>> >
>> >>>>> >> >>> >On Monday, February 24, 2025 at 7:34:13 PM UTC-8 Dima 
>> >>>>> >> >>> >Pasechnik wrote:
>> >>>>> >> >>> >
>> >>>>> >> >>> >> On Mon, Feb 24, 2025 at 5:44 PM John H Palmieri 
>> >>>>> >> >>> >> <[email protected]>
>> >>>>> >> >>> >> wrote:
>> >>>>> >> >>> >> >
>> >>>>> >> >>> >> > On OS X 15.3.1 with many homebrew packages installed, 
>> >>>>> >> >>> >> > including both
>> >>>>> >> >>> >> Python 3.12 and Python 3.13:
>> >>>>> >> >>> >> >
>> >>>>> >> >>> >> > - if I use Python 3.13, `pillow` and `cffi` fail to build. 
>> >>>>> >> >>> >> > (I used
>> >>>>> >> >>> >> "./configure", and I also tried after applying the patch 
>> >>>>> >> >>> >> from #39570)
>> >>>>> >> >>> >>
>> >>>>> >> >>> >> how about using installing pillow and cffi in Homebrew, and 
>> >>>>> >> >>> >> doing
>> >>>>> >> >>> >>
>> >>>>> >> >>> >> ./configure --enable-system-site-packages
>> >>>>> >> >>> >>
>> >>>>> >> >>> >> This should pick them from Homebrew. (If it doesn't it's a 
>> >>>>> >> >>> >> Sage bug I
>> >>>>> >> >>> >> suppose)
>> >>>>> >> >>> >>
>> >>>>> >> >>> >> Frankly speaking I am less than excited to try to figure out 
>> >>>>> >> >>> >> what goes
>> >>>>> >> >>> >> wrong in sdh_pip_install
>> >>>>> >> >>> >> in this case. Homebrew can do it...
>> >>>>> >> >>> >>
>> >>>>> >> >>> >> > - if I use Python 3.12, then `pillow` and `cffi` and 
>> >>>>> >> >>> >> > indeed everything
>> >>>>> >> >>> >> builds successfully. (I used "./configure 
>> >>>>> >> >>> >> --with-python=`which python3.12`")
>> >>>>> >> >>> >> >
>> >>>>> >> >>> >> >
>> >>>>> >> >>> >> > On Friday, February 21, 2025 at 3:40:17 PM UTC-8 Volker 
>> >>>>> >> >>> >> > Braun wrote:
>> >>>>> >> >>> >> >>
>> >>>>> >> >>> >> >> As always, you can get the latest beta version from the 
>> >>>>> >> >>> >> >> "develop" git
>> >>>>> >> >>> >> branch. Alternatively, the self-contained source tarball is 
>> >>>>> >> >>> >> at
>> >>>>> >> >>> >> http://www.sagemath.org/download-latest.html
>> >>>>> >> >>> >> >>
>> >>>>> >> >>> >> >>
>> >>>>> >> >>> >> >> 9cd86e9596a (tag: 10.6.beta7, github/develop) Updated 
>> >>>>> >> >>> >> >> SageMath version
>> >>>>> >> >>> >> to 10.6.beta7
>> >>>>> >> >>> >> >> c2766d1f625 gh-39541: Remove dead mailing lists.
>> >>>>> >> >>> >> >> 54951498178 gh-39536: Fix a nonfunctional long time 
>> >>>>> >> >>> >> >> doctest tag
>> >>>>> >> >>> >> >> 5289298248a gh-39533: build/pkgs: update eclib to version 
>> >>>>> >> >>> >> >> 20250122
>> >>>>> >> >>> >> >> 7420a272382 gh-39530: Improvement to flint_autogen reader
>> >>>>> >> >>> >> >> cee1e2b52f9 gh-39527: Test on CI that update-meson is 
>> >>>>> >> >>> >> >> properly ran
>> >>>>> >> >>> >> >> a007379b910 gh-39523: multi polynomial element/repr
>> >>>>> >> >>> >> >> 850493472d5 gh-39521: rename also set-like species
>> >>>>> >> >>> >> >> bd626efc02f gh-39518: line_graph for multigraphs
>> >>>>> >> >>> >> >> 72c6188116a gh-39513: Show test failures of ci-meson as 
>> >>>>> >> >>> >> >> annotations
>> >>>>> >> >>> >> >> 4f4c1203415 gh-39510: add of the function 
>> >>>>> >> >>> >> >> rank_support_of_vector
>> >>>>> >> >>> >> >> 2af3de6fa6e gh-39509: Add a doctest for
>> >>>>> >> >>> >> HomsetsCategory._make_named_class_key
>> >>>>> >> >>> >> >> d8b7cdfcecb gh-39508: Fix some typo
>> >>>>> >> >>> >> >> b567299fbbb gh-39506: New algorithm for cuts of a Poset
>> >>>>> >> >>> >> >> 88672d71a29 gh-39505: Improve handling of strings 
>> >>>>> >> >>> >> >> supplied as matrix
>> >>>>> >> >>> >> entries
>> >>>>> >> >>> >> >> 935232df552 gh-39504: Implement im_gens in the class 
>> >>>>> >> >>> >> >> fraction_field_FpT
>> >>>>> >> >>> >> >> a01815fda00 gh-39499: Improve sage_getfile by looking at 
>> >>>>> >> >>> >> >> __init__
>> >>>>> >> >>> >> >> b7da8f23c43 gh-39498: Apply sort and filter of 
>> >>>>> >> >>> >> >> walk_packages
>> >>>>> >> >>> >> consistently
>> >>>>> >> >>> >> >> 3500f616037 gh-39497: moving random_element to category 
>> >>>>> >> >>> >> >> of rings
>> >>>>> >> >>> >> >> e368fa71812 gh-39496: Add deformation cones and checking 
>> >>>>> >> >>> >> >> for regularity
>> >>>>> >> >>> >> for Point Configurations and normal fans of Polyhedra
>> >>>>> >> >>> >> >> 8978ff83f52 gh-39495: Add documentation of Gabidulin 
>> >>>>> >> >>> >> >> codes in the
>> >>>>> >> >>> >> reference manual
>> >>>>> >> >>> >> >> 98eec5d86fe gh-39494: Fix more doctests in meson_editable 
>> >>>>> >> >>> >> >> install
>> >>>>> >> >>> >> >> cfbaaf16be2 gh-39491: fixing doctests failures in 
>> >>>>> >> >>> >> >> misc/latex*.py
>> >>>>> >> >>> >> >> a2737325c19 gh-39489: fix several oeis related doctests
>> >>>>> >> >>> >> >> 8a44c7365b7 gh-39488: Fix issue on matrix construction 
>> >>>>> >> >>> >> >> over integer mod
>> >>>>> >> >>> >> ring for large coefficients
>> >>>>> >> >>> >> >> 78a3eede92c gh-39487: Improvement of the rendering of the 
>> >>>>> >> >>> >> >> documentation
>> >>>>> >> >>> >> in polynomial sequence
>> >>>>> >> >>> >> >> 34b7f9050fb gh-39486: Finish changing Rational's round 
>> >>>>> >> >>> >> >> method default
>> >>>>> >> >>> >> rounding to even
>> >>>>> >> >>> >> >> 04afcf71552 gh-39485: Implement conversion from laurent 
>> >>>>> >> >>> >> >> series to
>> >>>>> >> >>> >> rational function field
>> >>>>> >> >>> >> >> ae158663847 gh-39482: fix one typo
>> >>>>> >> >>> >> >> 83f7d065286 gh-39481: Fixed issue in list_plot where it 
>> >>>>> >> >>> >> >> assumed data
>> >>>>> >> >>> >> had been enumerated when it might not have been
>> >>>>> >> >>> >> >> 8e204dde3b3 gh-39479: Fixed crash when exp(0) of p-adic 
>> >>>>> >> >>> >> >> numbers is
>> >>>>> >> >>> >> called
>> >>>>> >> >>> >> >> 11ee7ca4e19 gh-39477: typing and details in tableaux files
>> >>>>> >> >>> >> >> c61e562de76 gh-39476: typing annotation for gens method 
>> >>>>> >> >>> >> >> in rings and
>> >>>>> >> >>> >> groups (pyx files)
>> >>>>> >> >>> >> >> 85ce297ff2c gh-39474: convert gens method in modular to 
>> >>>>> >> >>> >> >> return tuple
>> >>>>> >> >>> >> >> c732c6552e4 gh-39473: add tuple typing to gens methods in 
>> >>>>> >> >>> >> >> algebras
>> >>>>> >> >>> >> >> 556d5f0eb80 gh-39466: remove some deprecations in groups
>> >>>>> >> >>> >> >> c22e48480c3 gh-39456: Trivial simplifications for arccos
>> >>>>> >> >>> >> >> 68744be391e gh-39454: remove deprecated stuff in 
>> >>>>> >> >>> >> >> permutation.py
>> >>>>> >> >>> >> >> 784ff00867a gh-39449: Magma padics
>> >>>>> >> >>> >> >> 897e722c74e gh-39423: Use import_module instead of 
>> >>>>> >> >>> >> >> find_spec
>> >>>>> >> >>> >> >> 870aba20786 gh-39397: details in braid groups
>> >>>>> >> >>> >> >> e6975bebe80 gh-39381: Allow system python 3.13
>> >>>>> >> >>> >> >> b57e33797cf gh-39366: Add documentation to LaurentSeries 
>> >>>>> >> >>> >> >> point to
>> >>>>> >> >>> >> accessors
>> >>>>> >> >>> >> >> 357f95e088c gh-39365: Allow coercion from Frac(QQ[x]) to
>> >>>>> >> >>> >> LaurentSeriesRing(QQ)
>> >>>>> >> >>> >> >> 041336691f6 gh-39266: add parameter immutable to 
>> >>>>> >> >>> >> >> generators in
>> >>>>> >> >>> >> `sage/graphs/digraph_generators.py` (part 2)
>> >>>>> >> >>> >> >> e0806107d60 gh-39251: Require Python 3.11 or newer; 
>> >>>>> >> >>> >> >> remove outdated
>> >>>>> >> >>> >> workarounds
>> >>>>> >> >>> >> >> ac8406fba6a gh-39248: introduce new apozeta polynomial 
>> >>>>> >> >>> >> >> for posets
>> >>>>> >> >>> >> >> 10edc54de0a gh-39215: Class polynomial for Drinfeld 
>> >>>>> >> >>> >> >> modules
>> >>>>> >> >>> >> >> 2e0d6d201bf gh-39214: Add keyword prec for exponential 
>> >>>>> >> >>> >> >> and logarithm of
>> >>>>> >> >>> >> Drinfeld modules
>> >>>>> >> >>> >> >> b5c26156430 gh-39212: Refactor period lattice
>> >>>>> >> >>> >> >> d0e72a61433 gh-39128: Add note about makeflags and ninja 
>> >>>>> >> >>> >> >> parallelism
>> >>>>> >> >>> >> >> 067ebbd0c8c gh-39093: Improve hack used in debug_options
>> >>>>> >> >>> >> >> a89c57b64f8 gh-39092: Remove erroneous member declaration 
>> >>>>> >> >>> >> >> in
>> >>>>> >> >>> >> farey_symbol
>> >>>>> >> >>> >> >> cc231ef0353 gh-39061: Fix some errors in documentation of 
>> >>>>> >> >>> >> >> cachefunc
>> >>>>> >> >>> >> >> f0581cbe397 gh-39025: Add reseed_rng option to p_iter_fork
>> >>>>> >> >>> >> >> 290b261bc4f gh-38986: Allow CRT_list() to be called with 
>> >>>>> >> >>> >> >> one argument
>> >>>>> >> >>> >> >> 0a278fc918e gh-38824: Turn some doctests in 
>> >>>>> >> >>> >> >> `ell_rational_field.py`
>> >>>>> >> >>> >> into long tests
>> >>>>> >> >>> >> >> c43cd23ee83 gh-38650: Add support for pseudomorphisms
>> >>>>> >> >>> >> >> 6e82ee51eb9 gh-38108: Implicit function solver for lazy 
>> >>>>> >> >>> >> >> series
>> >>>>> >> >>> >> >> 42d00ed27b3 gh-37173: Implemented `.ramified_places` and 
>> >>>>> >> >>> >> >> modified
>> >>>>> >> >>> >> further methods to extend quaternion algebra functionality 
>> >>>>> >> >>> >> to number fields
>> >>>>> >> >>> >> >> 8d7107c1aab gh-37158: use Parent in quotient rings too
>> >>>>> >> >>> >> >> 766c7a0c5b8 (tag: 10.6.beta6) Updated SageMath version to 
>> >>>>> >> >>> >> >> 10.6.beta6
>> >>>>> >> >>> >> >
>> >>>>> >> >>> >> > --
>> >>>>> >> >>> >> > You received this message because you are subscribed to 
>> >>>>> >> >>> >> > the Google
>> >>>>> >> >>> >> Groups "sage-release" group.
>> >>>>> >> >>> >> > To unsubscribe from this group and stop receiving emails 
>> >>>>> >> >>> >> > from it, send
>> >>>>> >> >>> >> an email to [email protected].
>> >>>>> >> >>> >> > To view this discussion visit
>> >>>>> >> >>> >> https://groups.google.com/d/msgid/sage-release/aa27984d-e9b6-4eb4-b4a1-a422672e26a6n%40googlegroups.com
>> >>>>> >> >>> >> .
>> >>>>> >> >>> >>
>> >>>>> >> >>> >
>> >>>>> >> >
>> >>>>> >> > --
>> >>>>> >> > You received this message because you are subscribed to the 
>> >>>>> >> > Google Groups "sage-release" group.
>> >>>>> >> > To unsubscribe from this group and stop receiving emails from it, 
>> >>>>> >> > send an email to [email protected].
>> >>>>> >> > To view this discussion visit 
>> >>>>> >> > https://groups.google.com/d/msgid/sage-release/2c938919-99f9-49e2-865e-6bcc011a0098n%40googlegroups.com.
>> >>>>> >
>> >>>>> > --
>> >>>>> > You received this message because you are subscribed to the Google 
>> >>>>> > Groups "sage-release" group.
>> >>>>> > To unsubscribe from this group and stop receiving emails from it, 
>> >>>>> > send an email to [email protected].
>> >>>>> > To view this discussion visit 
>> >>>>> > https://groups.google.com/d/msgid/sage-release/c631ea8b-7c97-433f-9aa6-bf39686332b7n%40googlegroups.com.
>> >
>> > --
>> > You received this message because you are subscribed to the Google Groups 
>> > "sage-release" group.
>> > To unsubscribe from this group and stop receiving emails from it, send an 
>> > email to [email protected].
>> > To view this discussion visit 
>> > https://groups.google.com/d/msgid/sage-release/93906b68-a7ce-40eb-a879-b2978dee1fb0n%40googlegroups.com.
>>
>> --
>> You received this message because you are subscribed to a topic in the 
>> Google Groups "sage-release" group.
>> To unsubscribe from this topic, visit 
>> https://groups.google.com/d/topic/sage-release/8u71odPIXD8/unsubscribe.
>> To unsubscribe from this group and all its topics, send an email to 
>> [email protected].
>> To view this discussion visit 
>> https://groups.google.com/d/msgid/sage-release/CAAWYfq2MxfW1_XoqtKCdv7Vnp0m4jVAdC68pTwCxAN9WvbN0-w%40mail.gmail.com.
>
> --
> You received this message because you are subscribed to the Google Groups 
> "sage-release" group.
> To unsubscribe from this group and stop receiving emails from it, send an 
> email to [email protected].
> To view this discussion visit 
> https://groups.google.com/d/msgid/sage-release/CALcZXRGRoDSP7DoU8p%3DhyBVxTTHAnB%3DpQzVsjK6hp26mp%2BHz2g%40mail.gmail.com.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-release" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion visit 
https://groups.google.com/d/msgid/sage-release/CAAWYfq0io6NvR3tzTr%2BVpEF1KzFyHzvF%3DXhSSE%3DcJFz3RBCSuw%40mail.gmail.com.

Reply via email to