On Tue, Jul 28, 2020 at 11:59:03PM -0400, George Koehler wrote: > This update might break sparc64. I don't have sparc64, but found a > bug in 20.4.24 on powerpc64 that might affect other 64-bit big-endian > arch (mips64 and sparc64). It's BROKEN-mips64, but the sparc64 snap > has packages of clic and maxima with ecl-16.1.3.p3.
> The ecl update has been stuck because it breaks net/clic (as Solene > said). In other ways, the update looks good to me. > ecl-20.4.24 works on macppc, has the same number of failures as amd64: > Did 245 tests (0 crashed), 17999 checks. > Pass: 17912 (100%) > Fail: 87 ( 0%) > reed-alert seems to work on macppc. > I didn't try maxima on macppc. For maxima on amd64, I replicated the > test results that Timo posted, and I can run xmaxima. > For the bug in 20.4.24 that broke my powerpc64, I need to write a bug > report for upstream. ecl_min calls ecl_symbol_value() before > init_all_symbols(); reads p = q->data.value too early; gets p == 2 on > most arch, p == 0x200000000 on 64-bit big-endian; but if (p & 3) == 0 > then ecl_min fails with SIGSEGV. (powerpc64 also needs a diff for > lang/boehm-gc that I intend to share later.) --George I gave building it a try, it died early on with: /usr/ports/pobj/ecl-20.4.24/ecl-20.4.24/build/ecl/ecl_atomics.h:39:4: error: #er ror "ECL needs AO_fetch_compare_and_swap or an equivalent" # error "ECL needs AO_fetch_compare_and_swap or an equivalent" --Kurt
