On Thu, 24 Mar 2016 00:38:45 +0100 Mathieu Lirzin <m...@gnu.org> wrote:
> Hello, > > this is the debugged version of a patch previously sent here: > > https://lists.gnu.org/archive/html/guix-devel/2015-11/msg00024.html > > It is possible to directly checkout 'origin/wip-check' branch to review these > patches. Thanks. In addition to the test status, I get messages like this when running `make check`: ----- ;;; note: source file /home/bavier/projects/guix/srfi/srfi-64.scm ;;; newer than compiled /gnu/store/4r7iv7lz7vbcvzdjilqpiflzkh8nff3b-profile/lib/guile/2.0/ccache/srfi/srfi-64.go PASS: tests/elpa.scm ./test-env: line 1: 19815 Terminated "/home/bavier/projects/guix/build/master/pre-inst-env" "/home/bavier/projects/guix/build/master/guix-daemon" --disable-chroot --substitute-urls="$GUIX_BINARY_SUBSTITUTE_URL" 2> /dev/null ----- It seems that we don't compile any of the srfi/srfi-* source? I don't recall seeing the "Terminated" messages before (anyone else see this?), but this patch to test-env.in seems to help: --- a/test-env.in +++ b/test-env.in @@ -100,7 +100,8 @@ then --substitute-urls="$GUIX_BINARY_SUBSTITUTE_URL" 2>/dev/null & daemon_pid=$! - trap "kill $daemon_pid ; rm -rf $NIX_STATE_DIR" EXIT + trap "kill $daemon_pid ; wait $daemon_pid 2>/dev/null ; + rm -rf $NIX_STATE_DIR" EXIT fi # Avoid issues that could stem from l10n, such as language/encoding The result output seems more informative. The counts include individual tests within each tests/*.scm, correct? ============================================================================ Testsuite summary for GNU Guix 0.9.1 ============================================================================ # TOTAL: 677 # PASS: 671 # SKIP: 6 # XFAIL: 0 # FAIL: 0 # XPASS: 0 # ERROR: 0 ============================================================================ I notice in the test-suite.log that logs are included from tests that are skipped. This seems appropriate, but it appears to also include PASS output from other tests in that suite (e.g. all of tests/gexp.log is included since two of the 49 tests are skipp). Is it possible to include output from only the skipped tests? I'm not familiar for the trs spec. Thanks for this work. I had considered reworking the test suites with Autoconf's AutoTest for better unity, but I think what you've done here is quite a bit more elegant with less disruption. `~Eric
