Created attachment 253298 [details] 1.10.5 patch Update lang/julia to 1.10.5.
Created attachment 253299 [details] correct patch to 1.10.5
Hi Hiroo, would you like be the maintainer?
(In reply to Zsolt Udvari from comment #2) OK, I will be the maintainer.
(In reply to Hiroo Ono from comment #1) Did you test it? As I see the DISTVERSIONSUFFIX isn't needed: https://github.com/JuliaLang/julia/tags
(In reply to Zsolt Udvari from comment #4) Yes I did. built it by poudriere testport -i and run julia command and issued some commands. It is not a detailed test, but... DISTVERSIONSUFFIX was there before I touched the port's Makefile (lang/julia 1.9.x and before). I did not try to remove it because it worked.
(In reply to Hiroo Ono from comment #5) Sorry, it's working. I didn't see any progress bar while fetch (I don't know why). I'll test your patch.
You should strip many libraries: Warning: 'share/julia/compiled/v1.10/LibUnwind_jll/CxrEE_pvTvc.so' is not stripped consider trying INSTALL_TARGET=install-strip or using ${STRIP_CMD} Warning: 'share/julia/compiled/v1.10/LibUnwind_jll/CxrEE_I0SgR.so' is not stripped consider trying INSTALL_TARGET=install-strip or using ${STRIP_CMD} Warning: 'share/julia/compiled/v1.10/MPFR_jll/NBMLS_pvTvc.so' is not stripped consider trying INSTALL_TARGET=install-strip or using ${STRIP_CMD} Warning: 'share/julia/compiled/v1.10/MPFR_jll/NBMLS_I0SgR.so' is not stripped consider trying INSTALL_TARGET=install-strip or using ${STRIP_CMD} Warning: 'share/julia/compiled/v1.10/Profile/nGhxz_I0SgR.so' is not stripped consider trying INSTALL_TARGET=install-strip or using ${STRIP_CMD} Warning: 'share/julia/compiled/v1.10/Profile/nGhxz_pvTvc.so' is not stripped consider trying INSTALL_TARGET=install-strip or using ${STRIP_CMD} Warning: 'share/julia/compiled/v1.10/GMP_jll/1Lisu_pvTvc.so' is not stripped consider trying INSTALL_TARGET=install-strip or using ${STRIP_CMD} Warning: 'share/julia/compiled/v1.10/GMP_jll/1Lisu_I0SgR.so' is not stripped consider trying INSTALL_TARGET=install-strip or using ${STRIP_CMD} ...
I ran 'make test' but there are many tests with errors. Could you please check it?
> You should strip many libraries I am modifying the Makefile to do this. > I ran 'make test' but there are many tests with errors. These tests are known to fail for some reason or other. For example: [LinearAlgebra/test/blas.jl:734] Error: no BLAS/LAPACK library loaded for cblas_ddot64_() -> We use openblas as BLAS library, but it does not implement this function which results to: Evaluated: 23.0 === 1.01855797974e-312 (Not calculated) [LinearAlgebra/test/lu.jl:303] Expression: String(take!(bf)) == "LinearAlgebra.LU{Float64, Matrix{Float64}, Vector{$(Int)}} Evaluated: "LinearAlgebra.LU{Float64, Matrix{Float64}, Vector{Int32}} -> In stdlib/v1.10/LinearAlgebra/src/lu.jl, LU is defined as: LU(factors::AbstractMatrix{T}, ipiv::AbstractVector{<:Integer}, info::BlasInt) But as openblas is LP64, USE_BLAS is set to 0, when Intel MKL is not used. https://github.com/JuliaLang/julia/pull/11433/files Should I patch the failed tests to match the expected behavior on our platform so that all the tests are carried?
(In reply to Hiroo Ono from comment #9) The explanation was not sufficient. When USE_BLAS is set to 0, BlasInt is Int32 (it is Int64 when USE_BLAS=1). So, Evaluated: "LinearAlgebra.LU{Float64, Matrix{Float64}, Vector{Int32}}" vs. Expected: "LinearAlgebra.LU{Float64, Matrix{Float64}, Vector{Int64}}" In this case, Int64 of Expected should be Int32 when USE_BLAS=0.
(In reply to Hiroo Ono from comment #9) If you want test (I suggest) should do something. Patch the test, skip problematic tests, etc.
Created attachment 253644 [details] revised patch to 1.10.5 Updated the patch. Revised binary strip code and skipped some tests: * Library version check that use sysetm (package) libraries. Their version do not match with the version of bundled libraries. * cholmod, spqr and umfpack test for SparseArrays. They need SuiteSparse but it is not built (USE_GPL_LIBS=0). Trying to build it requires CUDA version of the libraries, which are not present on FreeBSD. * Some backtrace check in stdlib/Test/test/runtests.jl and test/errorshow.jl. They do not seem to match the expected, but inserting some println's sometimes change the content of backtrace. They seem to be unstable.
(In reply to Hiroo Ono from comment #12) Thanks. There are still errors in tests (14 fails and 7 errors), the 'file' tests contains 11 fails and 2 errors. Another thing: the patch contains many lint-specific errors and warnings. Could you please check the modified port with 'portlint'? This is part of ports-mgmt/portlint.
Friendly ping.
(In reply to Zsolt Udvari from comment #14) Currently, there are 3 curl failures that I fail to find the definitions for the actions.
(In reply to Hiroo Ono from comment #15) Could you please your latest patch?
(In reply to Zsolt Udvari from comment #16) *provide
No feedback.
^Triage: this does not appear to be "In Progress".