The memory statistics code leads to segfaults during initialization (on machines with InfiniPath networking): (gdb) bt full #0 ips_ptl_init (ep=0x1fc6af8, ptl=0x1fc6f88, ctl=0x1fc6d78) at ptl.c:224 err = PSM_OK num_of_send_bufs = 1024 num_of_send_desc = 4096 imm_size = 128 context = 0x1fc6b70 user_info = 0x1fc6b90 enable_shcontexts = 0 current_count = #1 0x00007fb2aa672abf in __psm_ep_open_internal ( unique_job_key=unique_job_key@entry=0x7ffed1ee5800 "<\207\020#5\271\267\200\354x\242e8\364zo", devid_enabled=devid_enabled@entry=0x7ffed1ee5724, opts_i=opts_i@entry=0x7ffed1ee5810, mq=, epo=epo@entry=0x7ffed1ee5710, epido=epido@entry=0x7ffed1ee5708) at psm_ep.c:929 ep = 0x1fc6af8 num_units = 1 len = err = epaddr = 0x1e9dd78 buf = "miriel044:2.0.", '\000' p = e = old_cpuaff = 0x0 old_unit = 0x0 yield_cnt = {e_void = 0xfa, e_str = 0xfa , e_int = 250, e_uint = 250, e_long = 250, e_ulong = 250, e_ulonglong = 250} no_cpuaff = {e_void = 0x0, e_str = 0x0, e_int = 0, e_uint = 0, e_long = 0, e_ulong = 0, e_ulonglong = 0} env_unit_id = {e_void = 0xffffffffffffffff, e_str = 0xffffffffffffffff , e_int = -1, e_uint = 4294967295, e_long = -1, e_ulong = 18446744073709551615, e_ulonglong = 18446744073709551615} env_port_id = {e_void = 0x0, e_str = 0x0, e_int = 0, e_uint = 0, e_long = 0, e_ulong = 0, e_ulonglong = 0} env_sl = {e_void = 0x0, e_str = 0x0, e_int = 0, e_uint = 0, e_long = 0, e_ulong = 0, e_ulonglong = 0} ptl_sizes = default_cpuaff = opts = {timeout = 180000000000, unit = -1, affinity = 0, shm_mbytes = 10, sendbufs_num = 1024, network_pkey = 65535, port = 0, outsl = 0, service_id = 1152940698815692800, path_res_type = PSM_PATH_RES_NONE, senddesc_num = 4096, imm_size = 128} amsh_ptl = 0x1fc6e48 ips_ptl = 0x1fc6f88 self_ptl = 0x1fc99c8 i = 3 It looks like ptl.c:24 is writing past the region that was malloc'd. Turning stats off solves the problem. diff --git a/psm_utils.c b/psm_utils.c index c8651fe..5514921 100644 --- a/psm_utils.c +++ b/psm_utils.c @@ -1058,7 +1058,7 @@ psmi_log_memstats(psmi_memtype_t type, int64_t nbytes) return; } -#define psmi_stats_mask PSMI_STATSTYPE_MEMORY +#define psmi_stats_mask 0 #ifdef malloc #undef malloc n title='2023-08-15 09:38:29 -0400'>2023-08-15refresh: Add --target-version option....* guix/scripts/refresh.scm (%options): Register 'target-version' long version. (update-specification->update-spec): Add a fallback-version argument. (options->update-specs): Honor target-version option. * tests/guix-refresh.sh: Test it. * doc/guix.texi (Invoking guix refresh): Document it. Reviewed-by: Ludovic Courtès <ludo@gnu.org> Maxim Cournoyer 2023-05-31upstream: Honor package properties for ignored and extra inputs....* guix/upstream.scm (update-package-inputs)[filtered-inputs] [regular-inputs, native-inputs, propagated-inputs]: New procedures. Use them in 'update-field' calls. * tests/guix-refresh.sh (GUIX_TEST_UPDATER_TARGETS): Add "libreoffice" to the dependencies of "the-test-package". Add 'updater-ignored-inputs' property to "the-test-package". * doc/guix.texi (Invoking guix refresh): Document it. Ludovic Courtès 2023-05-31upstream: 'update-package-source' edits input fields....Previously, 'guix refresh r-ggplot2 -u' and similar commands would print of list of input changes that would have to be made manually. With this change, 'guix refresh -u' takes care of updating input fields automatically. * guix/upstream.scm (update-package-inputs): New procedure. (update-package-source): Call it when 'upstream-source-inputs' returns true. * guix/scripts/refresh.scm (update-package): Remove iteration over the result of 'changed-inputs'. * guix/import/test.scm (available-updates): Add support for input lists. * tests/guix-refresh.sh (GUIX_TEST_UPDATER_TARGETS): Add input list for "the-test-package". Make sure 'guix refresh -u' updates 'inputs' accordingly. * doc/guix.texi (Invoking guix refresh): Mention it. Ludovic Courtès 2023-04-21tests: Relax two tests that expected a non-zero error code....* tests/guix-package-aliases.sh: "guix upgrade foo bar" has always returned zero; adjust accordingly. * tests/guix-refresh.sh: "guix refresh -t test idutils" and similar return zero; adjust accordingly. Ludovic Courtès 2023-04-21tests: Fix checks for expected failures....Addresses <https://issues.guix.gnu.org/62406>. With 'set -e', a return status inverted with '!' does not cause the shell to exit immediately. Instead use '&& false' to indicate an expected failure. * tests/guix-archive.sh, tests/guix-build-branch.sh, tests/guix-build.sh, tests/guix-daemon.sh, tests/guix-download.sh, tests/guix-environment-container.sh, tests/guix-environment.sh, tests/guix-gc.sh, tests/guix-git-authenticate.sh, tests/guix-graph.sh, tests/guix-hash.sh, tests/guix-home.sh, tests/guix-pack-relocatable.sh, tests/guix-pack.sh, tests/guix-package-aliases.sh, tests/guix-package-net.sh, tests/guix-package.sh, tests/guix-refresh.sh, tests/guix-shell.sh, tests/guix-style.sh, tests/guix-system.sh: Replace uses of '! ...' with '... && false' or `test ! ...` as appropriate. Signed-off-by: Ludovic Courtès <ludo@gnu.org> Eric Bavier 2023-01-25tests: Do not try to create "$HOME/.config"....Previously this test would fail when HOME=/homeless-shelter and said directory cannot be created. * tests/guix-refresh.sh: Pass '--keyring' together with '-u'. Ludovic Courtès 2023-01-08refresh: Add CLI tests....* guix/import/test.scm, tests/guix-refresh.sh: New files. * Makefile.am (MODULES, SH_TESTS): Add them. Ludovic Courtès