From: Vladislav Shpilevoy <v.shpilevoy@tarantool.org>
To: sergeyb@tarantool.org, tarantool-patches@dev.tarantool.org
Subject: Re: [Tarantool-patches] [PATCH v2 4/3] test: add additional checks
Date: Thu, 11 Jun 2020 20:42:05 +0200 [thread overview]
Message-ID: <f88e179f-307b-510f-e6db-5b6c852c7527@tarantool.org> (raw)
In-Reply-To: <cover.1591631501.git.sergeyb@tarantool.org>
Thanks for the patch! You didn't send it, so I did. I pasted it
below.
See 6 comments below.
>
> With diff-based approach test oracles were in .result files.
> After removing .result files test should be self-contained and
> shoud include strict checks to detect wrong behaviour during test
> execution.
>
> This commit brings additional test checks and allow diag() to fail a test.
>
> Part of #5000
>> diff --git a/test/unit/bloom.cc b/test/unit/bloom.cc
> index 7bd6814cc..7f91d97f7 100644
> --- a/test/unit/bloom.cc
> +++ b/test/unit/bloom.cc
> @@ -46,8 +46,8 @@ simple_test()
> if (fp_rate > p + 0.001)
> fp_rate_too_big++;
> }
> - note("error_count = %u", error_count);
> - note("fp_rate_too_big = %u", fp_rate_too_big);
> + is(error_count, 0, "error_count = %u", error_count);
> + is(fp_rate_too_big, 0, "fp_rate_too_big = %u", fp_rate_too_big);
> footer();
> }
>
> @@ -94,15 +94,15 @@ store_load_test()
> if (fp_rate > p + 0.001)
> fp_rate_too_big++;
> }
> - note("error_count = %d", error_count);
> - note("fp_rate_too_big = %d", fp_rate_too_big);
> + is(error_count, 0, "error_count = %u", error_count);
> + is(fp_rate_too_big, 0, "fp_rate_too_big = %u", fp_rate_too_big);
> footer();
> }
>
> int
> main(void)
> {
> - plan(0);
> + plan(4);
1. Normally, if a test case consists of multiple checks, it has
its own header()/footer() + plan()/check_plan(). So here the
plan should be 2. Simple_test() and store_load_test() should
have their own header()/plan()/check_plan()/footer().
> simple_test();
> store_load_test();
> check_plan();
> diff --git a/test/unit/cbus_hang.c b/test/unit/cbus_hang.c
> index eac283890..458d79de6 100644
> --- a/test/unit/cbus_hang.c
> +++ b/test/unit/cbus_hang.c
> @@ -148,8 +148,8 @@ main_f(va_list ap)
> cbus_stop_loop(&pipe_from_main_to_hang);
> cpipe_destroy(&pipe_from_main_to_hang);
>
> - cord_join(&hang_worker);
> - ok(true, "The hang worker has been joined");
> + int rc = cord_join(&hang_worker);
> + ok(!rc, "The hang worker has been joined");
2. Please, use 'is(rc, 0, ...)'. We don't use integers
as booleans. This is a part of the code style.
> alarm(0);
>
> diff --git a/test/unit/csv.c b/test/unit/csv.c
> index b79bf8f39..32c927327 100644
> --- a/test/unit/csv.c
> +++ b/test/unit/csv.c
3. This test is marked as 'pass' in test-run, but when I run it
manually, this is what I get:
Test failed: csv_get_error_status(&csv) == CSV_ER_INVALID is true at tarantool/test/unit/csv.c:154, in function 'test6'
> diff --git a/test/unit/fiber.cc b/test/unit/fiber.cc
> index 8568346e8..365a7c17b 100644
> --- a/test/unit/fiber.cc
> +++ b/test/unit/fiber.cc
> @@ -75,19 +75,26 @@ fiber_join_test()
> {
> header();
>
> - struct fiber *fiber = fiber_new_xc("join", noop_f);
> + int rc;
> + struct fiber *fiber = NULL;
> + fiber = fiber_new_xc("join", noop_f);
> + fail_if(!fiber);
4. Please, don't use '!' with non-boolean values. If you want
to check a pointer is not NULL, use != NULL. If you want to
check an integer is not 0, use != 0. I suggest you to find and
fix all the other similar places by yourself.
> fiber_set_joinable(fiber, true);
> fiber_wakeup(fiber);
> - fiber_join(fiber);
> + rc = fiber_join(fiber);
> + fail_if(rc);
>
> @@ -186,7 +203,7 @@ fiber_name_test()
> static int
> main_f(va_list ap)
> {
> - plan(0);
> + plan(2);
5. The same as the comment about bloom.cc.
> fiber_name_test();
> fiber_join_test();
> fiber_stack_test();
> @@ -200,7 +217,9 @@ int main()
> memory_init();
> fiber_init(fiber_cxx_invoke);
> fiber_attr_create(&default_attr);
> - struct fiber *main = fiber_new_xc("main", main_f);
> + struct fiber *main = NULL;
> + main = fiber_new_xc("main", main_f);
> + fail_if(!main);
> fiber_wakeup(main);
> ev_run(loop(), 0);
> fiber_free();
> diff --git a/test/unit/scramble.c b/test/unit/scramble.c
> index 65270bb4c..296a08715 100644
> --- a/test/unit/scramble.c
> +++ b/test/unit/scramble.c
6. I added abort() in the middle of the test. I crashes, and test-run
happily reports 'pass'. So if any of our unit tests will now crash,
we won't notice that (at least when plan is 0).
next prev parent reply other threads:[~2020-06-11 18:42 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-06-08 15:58 [Tarantool-patches] [PATCH v2 0/3] Make unit tests TAP-compliant sergeyb
2020-06-08 15:58 ` [Tarantool-patches] [PATCH v2 1/3] test: update unit test lib to produce TAP-compliant output sergeyb
2020-06-08 23:55 ` Vladislav Shpilevoy
2020-06-09 18:25 ` Sergey Bronnikov
2020-06-08 15:58 ` [Tarantool-patches] [PATCH v2 2/3] test: update unit tests to make them TAP-compliant sergeyb
2020-06-08 23:55 ` Vladislav Shpilevoy
2020-06-09 16:04 ` Sergey Bronnikov
2020-06-11 18:42 ` Vladislav Shpilevoy
2020-06-08 15:58 ` [Tarantool-patches] [PATCH v2 3/3] msgpuck: bump version to make test output TAP-compliant sergeyb
2020-06-08 16:03 ` [Tarantool-patches] [PATCH v2 4/3][msgpuck] test: " Sergey Bronnikov
2020-06-11 18:42 ` Vladislav Shpilevoy [this message]
2020-06-11 18:42 ` [Tarantool-patches] [PATCH v2 0/3] Make unit tests TAP-compliant Vladislav Shpilevoy
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=f88e179f-307b-510f-e6db-5b6c852c7527@tarantool.org \
--to=v.shpilevoy@tarantool.org \
--cc=sergeyb@tarantool.org \
--cc=tarantool-patches@dev.tarantool.org \
--subject='Re: [Tarantool-patches] [PATCH v2 4/3] test: add additional checks' \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox