Relying on meson_src_test may end up with limited src_test output in build log. Compare e.g. x11-libs/gdk-pixbuf with current meson_src_test outputs: >>> Test phase: x11-libs/gdk-pixbuf-2.38.1 * abi_x86_64.amd64: running multilib-minimal_abi_src_test ninja -v -j7 -l15 -C /tmp/portage/x11-libs/gdk-pixbuf-2.38.1/work/gdk-pixbuf-2.38.1-abi_x86_64.amd64 test ninja: Entering directory `/tmp/portage/x11-libs/gdk-pixbuf-2.38.1/work/gdk-pixbuf-2.38.1-abi_x86_64.amd64' [0/1] /usr/lib64/python-exec/python3.6/meson test --no-rebuild --print-errorlogs 1/22 gdk-pixbuf:format / animation OK 0.04 s 2/22 gdk-pixbuf:security / cve-2015-4491 OK 0.93 s 3/22 gdk-pixbuf:conform+slow / pixbuf-fail OK 3.26 s 4/22 gdk-pixbuf:conform / pixbuf-icon-serialize OK 0.05 s 5/22 gdk-pixbuf:slow / pixbuf-randomly-modified OK 1.72 s 6/22 gdk-pixbuf:io / pixbuf-threads OK 0.06 s 7/22 gdk-pixbuf:io / pixbuf-gif OK 0.13 s 8/22 gdk-pixbuf:io / pixbuf-icc OK 0.04 s 9/22 gdk-pixbuf:format / pixbuf-jpeg OK 0.05 s 10/22 gdk-pixbuf:io / pixbuf-dpi OK 0.03 s 11/22 gdk-pixbuf:format / pixbuf-pixdata OK 0.03 s 12/22 gdk-pixbuf:io / pixbuf-stream OK 0.03 s 13/22 gdk-pixbuf:conform / pixbuf-reftest OK 0.40 s 14/22 gdk-pixbuf:io / pixbuf-resource OK 0.02 s 15/22 gdk-pixbuf:ops / pixbuf-scale OK 0.10 s 16/22 gdk-pixbuf:ops / pixbuf-scale-two-step OK 0.14 s 17/22 gdk-pixbuf:format / pixbuf-short-gif-write OK 0.11 s 18/22 gdk-pixbuf:format / pixbuf-gif-circular-table OK 0.04 s 19/22 gdk-pixbuf:io / pixbuf-save OK 0.03 s 20/22 gdk-pixbuf:conform / pixbuf-readonly-to-mutable OK 0.02 s 21/22 gdk-pixbuf:ops / pixbuf-composite OK 0.02 s 22/22 gdk-pixbuf:slow / pixbuf-area-updated OK 78.99 s OK: 22 FAIL: 0 SKIP: 0 TIMEOUT: 0 Full log written to /tmp/portage/x11-libs/gdk-pixbuf-2.38.1/work/gdk-pixbuf-2.38.1-abi_x86_64.amd64/meson-logs/testlog.txt >>> Completed testing x11-libs/gdk-pixbuf-2.38.1 while meson test -C ${BUILD_DIR} -v || die "tests failed" outputs: >>> Test phase: x11-libs/gdk-pixbuf-2.38.1 * abi_x86_64.amd64: running multilib-minimal_abi_src_test ninja: Entering directory `/tmp/portage/x11-libs/gdk-pixbuf-2.38.1/work/gdk-pixbuf-2.38.1-abi_x86_64.amd64' ninja: no work to do. # random seed: R02S142359ca28b763c450b4064c40ff9286 1..2 # Start of animation tests ok 1 /animation/gif ok 2 /animation/ani # End of animation tests <snip> 20/22 gdk-pixbuf:conform / pixbuf-readonly-to-mutable OK 0.03 s 21/22 gdk-pixbuf:ops / pixbuf-composite OK 0.05 s 22/22 gdk-pixbuf:slow / pixbuf-area-updated OK 78.35 s OK: 22 FAIL: 0 SKIP: 0 TIMEOUT: 0 >>> Completed testing x11-libs/gdk-pixbuf-2.38.1 I guess with the <snip> it's harder to see, but it's much more verbose, and the small version only is outputted after all is done too, it seems, so don't even see all the tests but (very slow) last are done until all are done. I suspect my meson test -C ${BUILD_DIR} -v is not ideal in that it probably ignored MAKEOPTS, so this is a bug report, not patch :( An additional issue is that meson_src_test can't be combined with running other programs, so then direct meson test or eninja calls are also needed, thus potentially also ending up ignoring MAKEOPTS due to ebuild not taking care of that. This is a common problem when tests need to be wrapped inside dbus-run-session; and things get even more complicated if virtx needs to be involved too.
Created attachment 571290 [details, diff] meson-fix-tests.patch What happens if you apply this patch to meson.eclass? This changes the way we run tests to use the meson test command.
Created attachment 571292 [details, diff] meson-fix-tests.patch This is an updated version which adds an emesontestargs variable which behaves like emesonargs.
Please don't add more variables like "emesontestargs". Just pass "$@" instead. Also, you have not quoted the variables correctly.
(In reply to Mike Gilbert from comment #3) > Please don't add more variables like "emesontestargs". Just pass "$@" > instead. I see cmake-utils has myctestargs, so I guess this is somewhat common. Let's just support both the variable and "$@".
Created attachment 571410 [details, diff] 0001-meson.eclass-fix-tests.patch Here is the latest iteration of this fix.
I fixed the lower case references to ninjaopts and makeopts. They are all upper case.
(In reply to William Hubbs from comment #5) > Created attachment 571410 [details, diff] [details, diff] > 0001-meson.eclass-fix-tests.patch > > Here is the latest iteration of this fix. There is an unnecessary tab in the middle of this line: + [[ -n ${ninjaopts} ]] || [[ -n ${makeopts} ]] && ^^^^^^^ (In reply to William Hubbs from comment #6) > I fixed the lower case references to ninjaopts and makeopts. They are > all upper case. You can also merge two [[ ]] into one: + [[ -n ${NINJAOPTS} || -n ${MAKEOPTS} ]] &&
Created attachment 577920 [details] 0001-meson.eclass-fix-tests.patch Here is the latest iteration of this fix.
The bug has been closed via the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=fe02c66d4b7638575778dcb7cbeb09c42a83c764 commit fe02c66d4b7638575778dcb7cbeb09c42a83c764 Author: William Hubbs <williamh@gentoo.org> AuthorDate: 2019-06-02 16:29:07 +0000 Commit: William Hubbs <williamh@gentoo.org> CommitDate: 2019-06-02 16:47:11 +0000 meson.eclass: fix tests - use "meson test" to run tests instead of "eninja test" - calculate --num-processes based on ninjaopts or makeopts in the "meson test" call. - allow meson_src_test to accept arguments with the emesontestargs array or passed in the call like meson_src_configure Closes: https://bugs.gentoo.org/680010 Signed-off-by: William Hubbs <williamh@gentoo.org> eclass/meson.eclass | 25 +++++++++++++++++++++++-- 1 file changed, 23 insertions(+), 2 deletions(-)
