From 8f87dd1eaa80bdad501ad6f7767f14b4e0cb01d6 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= Date: Tue, 30 May 2017 09:40:43 +0200 Subject: Fix process_SUITE:low_prio/1 It is expected that low-priority processes should be allowed to run about 1/8 of the time of normal-priority processes. The proportion calculated was usually considerably lower than 1/8. It seems that the main reason is the punishement in reductions when sending to a process that alreday has many messages in the message queue. I have verified that by running the emulator with the +snsp to turn off the send punishment. To avoid that punishment, let the server process run at high priority so that it can keep its message queue short. It seems that the proportion is now very close to 1/8. Therefore, we can tighten the margins for the proportion. --- erts/emulator/test/process_SUITE.erl | 50 ++++++++++++++++++++++-------------- 1 file changed, 31 insertions(+), 19 deletions(-) diff --git a/erts/emulator/test/process_SUITE.erl b/erts/emulator/test/process_SUITE.erl index e14185e881..a65c3f0cad 100644 --- a/erts/emulator/test/process_SUITE.erl +++ b/erts/emulator/test/process_SUITE.erl @@ -1024,36 +1024,48 @@ bump_big(Prev, Limit) -> %% Priority 'low' should be mixed with 'normal' using a factor of %% about 8. (OTP-2644) low_prio(Config) when is_list(Config) -> - case erlang:system_info(schedulers_online) of - 1 -> - ok = low_prio_test(Config); - _ -> - erlang:system_flag(multi_scheduling, block_normal), - ok = low_prio_test(Config), - erlang:system_flag(multi_scheduling, unblock_normal), - {comment, - "Test not written for SMP runtime system. " - "Multi scheduling blocked during test."} - end. + erlang:system_flag(multi_scheduling, block_normal), + Prop = low_prio_test(Config), + erlang:system_flag(multi_scheduling, unblock_normal), + Str = lists:flatten(io_lib:format("Low/high proportion is ~.3f", + [Prop])), + {comment,Str}. low_prio_test(Config) when is_list(Config) -> process_flag(trap_exit, true), - S = spawn_link(?MODULE, prio_server, [0, 0]), + + %% Spawn the server running with high priority. The server must + %% not run at normal priority as that would skew the results for + %% two reasons: + %% + %% 1. There would be one more normal-priority processes than + %% low-priority processes. + %% + %% 2. The receive queue would grow faster than the server process + %% could process it. That would in turn trigger the reduction + %% punishment for the clients. + S = spawn_opt(?MODULE, prio_server, [0, 0], [link,{priority,high}]), + + %% Spawn the clients and let them run for a while. PCs = spawn_prio_clients(S, erlang:system_info(schedulers_online)), - ct:sleep({seconds,3}), + ct:sleep({seconds,2}), lists:foreach(fun (P) -> exit(P, kill) end, PCs), + + %% Stop the server and retrieve the result. S ! exit, - receive {'EXIT', S, {A, B}} -> check_prio(A, B) end, - ok. + receive + {'EXIT', S, {A, B}} -> + check_prio(A, B) + end. check_prio(A, B) -> Prop = A/B, ok = io:format("Low=~p, High=~p, Prop=~p\n", [A, B, Prop]), - %% It isn't 1/8, it's more like 0.3, but let's check that - %% the low-prio processes get some little chance to run at all. - true = (Prop < 1.0), - true = (Prop > 1/32). + %% Prop is expected to be appr. 1/8. Allow a reasonable margin. + true = Prop < 1/4, + true = Prop > 1/16, + Prop. prio_server(A, B) -> receive -- cgit v1.2.3 From 8381b753ce94769691a6dc499f4d40373cb0504c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= Date: Tue, 30 May 2017 14:05:35 +0200 Subject: Lengthen too short timetraps c062dfc485a added timetraps to test suites that had no timetraps. Some of added timetraps are only 5 or 10 seconds, which is can be too short time on some slow computers. Change those times to one or two minutes. --- erts/emulator/test/bs_construct_SUITE.erl | 2 +- erts/emulator/test/call_trace_SUITE.erl | 2 +- erts/emulator/test/ddll_SUITE.erl | 2 +- erts/emulator/test/evil_SUITE.erl | 2 +- erts/emulator/test/exception_SUITE.erl | 2 +- erts/emulator/test/lttng_SUITE.erl | 2 +- erts/emulator/test/match_spec_SUITE.erl | 2 +- erts/emulator/test/nested_SUITE.erl | 2 +- erts/emulator/test/port_SUITE.erl | 2 +- erts/emulator/test/port_trace_SUITE.erl | 2 +- erts/emulator/test/trace_SUITE.erl | 2 +- erts/emulator/test/trace_port_SUITE.erl | 2 +- 12 files changed, 12 insertions(+), 12 deletions(-) diff --git a/erts/emulator/test/bs_construct_SUITE.erl b/erts/emulator/test/bs_construct_SUITE.erl index ed03284a5b..779d81daa5 100644 --- a/erts/emulator/test/bs_construct_SUITE.erl +++ b/erts/emulator/test/bs_construct_SUITE.erl @@ -35,7 +35,7 @@ suite() -> [{ct_hooks,[ts_install_cth]}, - {timetrap, {seconds, 10}}]. + {timetrap, {minutes, 1}}]. all() -> [test1, test2, test3, test4, test5, testf, not_used, diff --git a/erts/emulator/test/call_trace_SUITE.erl b/erts/emulator/test/call_trace_SUITE.erl index 95171d04ce..1251d644ae 100644 --- a/erts/emulator/test/call_trace_SUITE.erl +++ b/erts/emulator/test/call_trace_SUITE.erl @@ -43,7 +43,7 @@ suite() -> [{ct_hooks,[ts_install_cth]}, - {timetrap, {seconds, 30}}]. + {timetrap, {minutes, 2}}]. all() -> Common = [errors, on_load], diff --git a/erts/emulator/test/ddll_SUITE.erl b/erts/emulator/test/ddll_SUITE.erl index 0b9f76a892..031b05790d 100644 --- a/erts/emulator/test/ddll_SUITE.erl +++ b/erts/emulator/test/ddll_SUITE.erl @@ -55,7 +55,7 @@ suite() -> [{ct_hooks,[ts_install_cth]}, - {timetrap, {seconds, 10}}]. + {timetrap, {minutes, 1}}]. all() -> [ddll_test, errors, reference_count, kill_port, diff --git a/erts/emulator/test/evil_SUITE.erl b/erts/emulator/test/evil_SUITE.erl index 9416ac7a02..fb1954ce37 100644 --- a/erts/emulator/test/evil_SUITE.erl +++ b/erts/emulator/test/evil_SUITE.erl @@ -34,7 +34,7 @@ suite() -> [{ct_hooks,[ts_install_cth]}, - {timetrap, {seconds, 30}}]. + {timetrap, {minutes, 1}}]. all() -> [heap_frag, encode_decode_ext, decode_integer_ext, diff --git a/erts/emulator/test/exception_SUITE.erl b/erts/emulator/test/exception_SUITE.erl index 76e3556bc4..ad6d8c890f 100644 --- a/erts/emulator/test/exception_SUITE.erl +++ b/erts/emulator/test/exception_SUITE.erl @@ -33,7 +33,7 @@ suite() -> [{ct_hooks,[ts_install_cth]}, - {timetrap, {seconds, 10}}]. + {timetrap, {minutes, 1}}]. all() -> [badmatch, pending_errors, nil_arith, stacktrace, diff --git a/erts/emulator/test/lttng_SUITE.erl b/erts/emulator/test/lttng_SUITE.erl index c12f63706a..1c1952f912 100644 --- a/erts/emulator/test/lttng_SUITE.erl +++ b/erts/emulator/test/lttng_SUITE.erl @@ -40,7 +40,7 @@ suite() -> [{ct_hooks,[ts_install_cth]}, - {timetrap, {seconds, 10}}]. + {timetrap, {minutes, 1}}]. all() -> [t_lttng_list, diff --git a/erts/emulator/test/match_spec_SUITE.erl b/erts/emulator/test/match_spec_SUITE.erl index eb189c2c33..92ddc23592 100644 --- a/erts/emulator/test/match_spec_SUITE.erl +++ b/erts/emulator/test/match_spec_SUITE.erl @@ -42,7 +42,7 @@ suite() -> [{ct_hooks,[ts_install_cth]}, - {timetrap, {seconds, 30}}]. + {timetrap, {minutes, 1}}]. all() -> case test_server:is_native(match_spec_SUITE) of diff --git a/erts/emulator/test/nested_SUITE.erl b/erts/emulator/test/nested_SUITE.erl index 7af2873ce2..5059317172 100644 --- a/erts/emulator/test/nested_SUITE.erl +++ b/erts/emulator/test/nested_SUITE.erl @@ -27,7 +27,7 @@ suite() -> [{ct_hooks,[ts_install_cth]}, - {timetrap, {seconds, 10}}]. + {timetrap, {minutes, 1}}]. all() -> [case_in_case, case_in_after, catch_in_catch, diff --git a/erts/emulator/test/port_SUITE.erl b/erts/emulator/test/port_SUITE.erl index 94ee9851dd..fccbaf13ee 100644 --- a/erts/emulator/test/port_SUITE.erl +++ b/erts/emulator/test/port_SUITE.erl @@ -153,7 +153,7 @@ suite() -> [{ct_hooks,[ts_install_cth]}, - {timetrap, {seconds, 10}}]. + {timetrap, {minutes, 1}}]. all() -> [otp_6224, {group, stream}, basic_ping, slow_writes, diff --git a/erts/emulator/test/port_trace_SUITE.erl b/erts/emulator/test/port_trace_SUITE.erl index 03efdc15db..bfc3c8cb51 100644 --- a/erts/emulator/test/port_trace_SUITE.erl +++ b/erts/emulator/test/port_trace_SUITE.erl @@ -52,7 +52,7 @@ -define(ECHO_DRV_REMOTE_SEND_TERM, 15). suite() -> [{ct_hooks,[ts_install_cth]}, - {timetrap, {seconds, 30}}]. + {timetrap, {minutes, 2}}]. all() -> [port_specs, ports, open_close, diff --git a/erts/emulator/test/trace_SUITE.erl b/erts/emulator/test/trace_SUITE.erl index 643c2e0472..53249229c9 100644 --- a/erts/emulator/test/trace_SUITE.erl +++ b/erts/emulator/test/trace_SUITE.erl @@ -46,7 +46,7 @@ suite() -> [{ct_hooks,[ts_install_cth]}, - {timetrap, {seconds, 5}}]. + {timetrap, {minutes, 1}}]. all() -> [cpu_timestamp, receive_trace, link_receive_call_correlation, diff --git a/erts/emulator/test/trace_port_SUITE.erl b/erts/emulator/test/trace_port_SUITE.erl index e4db368ea1..5eb27a7b68 100644 --- a/erts/emulator/test/trace_port_SUITE.erl +++ b/erts/emulator/test/trace_port_SUITE.erl @@ -37,7 +37,7 @@ suite() -> [{ct_hooks,[ts_install_cth]}, - {timetrap, {seconds, 30}}]. + {timetrap, {minutes, 2}}]. all() -> [call_trace, return_trace, send, receive_trace, -- cgit v1.2.3 From 98fde55efd0f8c3f5013329cf3d1acfe7516ab38 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= Date: Tue, 30 May 2017 14:25:34 +0200 Subject: receive_SUITE: Remove ?line macros --- erts/emulator/test/trace_SUITE.erl | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/erts/emulator/test/trace_SUITE.erl b/erts/emulator/test/trace_SUITE.erl index 53249229c9..bd0ea22de9 100644 --- a/erts/emulator/test/trace_SUITE.erl +++ b/erts/emulator/test/trace_SUITE.erl @@ -184,10 +184,10 @@ receive_trace(Config) when is_list(Config) -> {'EXIT', Intruder, {badarg, _}} = receive_first(), %% Untrace the process; we should not receive anything. - ?line 1 = erlang:trace(Receiver, false, ['receive']), - ?line Receiver ! {hello, there}, - ?line Receiver ! any_garbage, - ?line receive_nothing(), + 1 = erlang:trace(Receiver, false, ['receive']), + Receiver ! {hello, there}, + Receiver ! any_garbage, + receive_nothing(), %% Verify restrictions in matchspec for 'receive' F3 = fun (Pat) -> {'EXIT', {badarg,_}} = (catch erlang:trace_pattern('receive', Pat, [])) end, -- cgit v1.2.3 From f5c6195f7cfaea66ab3137ba5fa30f90626c8213 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= Date: Tue, 30 May 2017 16:14:32 +0200 Subject: distribution_SUITE: Refactor bulk_send_bigbig/1 Refactor bulk_send_bigbig/1 to make it somewhat easier to follow. --- erts/emulator/test/distribution_SUITE.erl | 122 +++++++++++++++--------------- 1 file changed, 59 insertions(+), 63 deletions(-) diff --git a/erts/emulator/test/distribution_SUITE.erl b/erts/emulator/test/distribution_SUITE.erl index 434e729310..78d67ef2ae 100644 --- a/erts/emulator/test/distribution_SUITE.erl +++ b/erts/emulator/test/distribution_SUITE.erl @@ -62,8 +62,7 @@ -export([sender/3, receiver2/2, dummy_waiter/0, dead_process/0, roundtrip/1, bounce/1, do_dist_auto_connect/1, inet_rpc_server/1, dist_parallel_sender/3, dist_parallel_receiver/0, - dist_evil_parallel_receiver/0, - sendersender/4, sendersender2/4]). + dist_evil_parallel_receiver/0]). %% epmd_module exports -export([start_link/0, register_node/2, register_node/3, port_please/2]). @@ -129,33 +128,43 @@ bulk_send_small(Config) when is_list(Config) -> bulk_send_big(Config) when is_list(Config) -> bulk_send(32, 64). -bulk_send_bigbig(Config) when is_list(Config) -> - bulk_sendsend(32*5, 4). - bulk_send(Terms, BinSize) -> ct:timetrap({seconds, 30}), io:format("Sending ~w binaries, each of size ~w K", [Terms, BinSize]), {ok, Node} = start_node(bulk_receiver), Recv = spawn(Node, erlang, apply, [fun receiver/2, [0, 0]]), - Bin = list_to_binary(lists:duplicate(BinSize*1024, 253)), + Bin = binary:copy(<<253>>, BinSize*1024), Size = Terms*size(Bin), {Elapsed, {Terms, Size}} = test_server:timecall(?MODULE, sender, [Recv, Bin, Terms]), stop_node(Node), - {comment, integer_to_list(trunc(Size/1024/max(1,Elapsed)+0.5)) ++ " K/s"}. + {comment, integer_to_list(round(Size/1024/max(1,Elapsed))) ++ " K/s"}. + +sender(To, _Bin, 0) -> + To ! {done, self()}, + receive + Any -> + Any + end; +sender(To, Bin, Left) -> + To ! {term, Bin}, + sender(To, Bin, Left-1). -bulk_sendsend(Terms, BinSize) -> +bulk_send_bigbig(Config) when is_list(Config) -> + Terms = 32*5, + BinSize = 4, {Rate1, MonitorCount1} = bulk_sendsend2(Terms, BinSize, 5), {Rate2, MonitorCount2} = bulk_sendsend2(Terms, BinSize, 995), Ratio = if MonitorCount2 == 0 -> MonitorCount1 / 1.0; true -> MonitorCount1 / MonitorCount2 end, - Comment = integer_to_list(Rate1) ++ " K/s, " ++ - integer_to_list(Rate2) ++ " K/s, " ++ - integer_to_list(MonitorCount1) ++ " monitor msgs, " ++ - integer_to_list(MonitorCount2) ++ " monitor msgs, " ++ - float_to_list(Ratio) ++ " monitor ratio", + Comment0 = io_lib:format("~p K/s, ~p K/s, " + "~p monitor msgs, ~p monitor msgs, " + "~.1f monitor ratio", + [Rate1,Rate2,MonitorCount1, + MonitorCount2,Ratio]), + Comment = lists:flatten(Comment0), if %% A somewhat arbitrary ratio, but hopefully one that will %% accommodate a wide range of CPU speeds. @@ -169,12 +178,11 @@ bulk_sendsend(Terms, BinSize) -> bulk_sendsend2(Terms, BinSize, BusyBufSize) -> ct:timetrap({seconds, 30}), - io:format("Sending ~w binaries, each of size ~w K", + io:format("\nSending ~w binaries, each of size ~w K", [Terms, BinSize]), {ok, NodeRecv} = start_node(bulk_receiver), Recv = spawn(NodeRecv, erlang, apply, [fun receiver/2, [0, 0]]), - Bin = list_to_binary(lists:duplicate(BinSize*1024, 253)), - %%Size = Terms*size(Bin), + Bin = binary:copy(<<253>>, BinSize*1024), %% SLF LEFT OFF HERE. %% When the caller uses small hunks, like 4k via @@ -185,74 +193,62 @@ bulk_sendsend2(Terms, BinSize, BusyBufSize) -> %% default busy size and "+zdbbl 5", and if the 5 case gets %% "many many more" monitor messages, then we know we're working. - {ok, NodeSend} = start_node(bulk_sender, "+zdbbl " ++ integer_to_list(BusyBufSize)), - _Send = spawn(NodeSend, erlang, apply, [fun sendersender/4, [self(), Recv, Bin, Terms]]), + {ok, NodeSend} = start_node(bulk_sender, "+zdbbl " ++ + integer_to_list(BusyBufSize)), + _Send = spawn(NodeSend, erlang, apply, + [fun sendersender/4, [self(), Recv, Bin, Terms]]), {Elapsed, {_TermsN, SizeN}, MonitorCount} = - receive - %% On some platforms (windows), the time taken is 0 so we - %% simulate that some little time has passed. - {sendersender, {0.0,T,MC}} -> - {0.0015, T, MC}; - {sendersender, BigRes} -> - BigRes - end, + receive + %% On some platforms (Windows), the time taken is 0 so we + %% simulate that some little time has passed. + {sendersender, {0.0,T,MC}} -> + {0.0015, T, MC}; + {sendersender, BigRes} -> + BigRes + end, stop_node(NodeRecv), stop_node(NodeSend), - {trunc(SizeN/1024/Elapsed+0.5), MonitorCount}. - -sender(To, _Bin, 0) -> - To ! {done, self()}, - receive - Any -> - Any - end; -sender(To, Bin, Left) -> - To ! {term, Bin}, - sender(To, Bin, Left-1). + {round(SizeN/1024/Elapsed), MonitorCount}. %% Sender process to be run on a slave node sendersender(Parent, To, Bin, Left) -> erlang:system_monitor(self(), [busy_dist_port]), - [spawn(fun() -> sendersender2(To, Bin, Left, false) end) || - _ <- lists:seq(1,1)], + _ = spawn(fun() -> + sendersender_send(To, Bin, Left), + exit(normal) + end), {USec, {Res, MonitorCount}} = - timer:tc(?MODULE, sendersender2, [To, Bin, Left, true]), + timer:tc(fun() -> + sendersender_send(To, Bin, Left), + To ! {done, self()}, + count_monitors(0) + end), Parent ! {sendersender, {USec/1000000, Res, MonitorCount}}. -sendersender2(To, Bin, Left, SendDone) -> - sendersender3(To, Bin, Left, SendDone, 0). +sendersender_send(_To, _Bin, 0) -> + ok; +sendersender_send(To, Bin, Left) -> + To ! {term, Bin}, + sendersender_send(To, Bin, Left-1). -sendersender3(To, _Bin, 0, SendDone, MonitorCount) -> - if SendDone -> - To ! {done, self()}; - true -> - ok - end, +count_monitors(MonitorCount) -> receive {monitor, _Pid, _Type, _Info} -> - sendersender3(To, _Bin, 0, SendDone, MonitorCount + 1) + count_monitors(MonitorCount + 1) after 0 -> - if SendDone -> - receive - Any when is_tuple(Any), size(Any) == 2 -> - {Any, MonitorCount} - end; - true -> - exit(normal) - end - end; -sendersender3(To, Bin, Left, SendDone, MonitorCount) -> - To ! {term, Bin}, - %%timer:sleep(50), - sendersender3(To, Bin, Left-1, SendDone, MonitorCount). + receive + {_,_}=Any -> + {Any,MonitorCount} + end + end. %% Receiver process to be run on a slave node. receiver(Terms, Size) -> receive {term, Bin} -> - receiver(Terms+1, Size+size(Bin)); + receiver(Terms+1, Size+byte_size(Bin)); {done, ReplyTo} -> ReplyTo ! {Terms, Size} end. -- cgit v1.2.3 From 94e8b18640304bb20dc591a459802ba99dbbd39e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= Date: Tue, 30 May 2017 16:18:10 +0200 Subject: Make bulk_send_bigbig/1 purely informational bulk_send_bigbig/1 sometimes fails even if there is nothing wrong. Stop testing the ratio, and just return information about the speed and number of monitor messages. The testcase will still serve as a smoke test of the +zdbbl option. --- erts/emulator/test/distribution_SUITE.erl | 10 +--------- 1 file changed, 1 insertion(+), 9 deletions(-) diff --git a/erts/emulator/test/distribution_SUITE.erl b/erts/emulator/test/distribution_SUITE.erl index 78d67ef2ae..dc7afa381b 100644 --- a/erts/emulator/test/distribution_SUITE.erl +++ b/erts/emulator/test/distribution_SUITE.erl @@ -165,15 +165,7 @@ bulk_send_bigbig(Config) when is_list(Config) -> [Rate1,Rate2,MonitorCount1, MonitorCount2,Ratio]), Comment = lists:flatten(Comment0), - if - %% A somewhat arbitrary ratio, but hopefully one that will - %% accommodate a wide range of CPU speeds. - Ratio > 8.0 -> - {comment,Comment}; - true -> - io:put_chars(Comment), - ct:fail(ratio_too_low) - end. + {comment,Comment}. bulk_sendsend2(Terms, BinSize, BusyBufSize) -> ct:timetrap({seconds, 30}), -- cgit v1.2.3 From 1bb171a7c27fcb79435a915bac834a4013b5f0dc Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= Date: Thu, 1 Jun 2017 07:29:23 +0200 Subject: Contain damage cause by spawn_opt_max_heap_size failing If process_SUITE:spawn_opt_max_heap_size/1 failed, the default value for the maximum heap size could have been changed. That would cause other test cases that spawned huge processes to fail. Contain the damage by always restoring the default value for max_heap_size in end_per_testcase/2. --- erts/emulator/test/process_SUITE.erl | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/erts/emulator/test/process_SUITE.erl b/erts/emulator/test/process_SUITE.erl index a65c3f0cad..180311202f 100644 --- a/erts/emulator/test/process_SUITE.erl +++ b/erts/emulator/test/process_SUITE.erl @@ -134,6 +134,11 @@ init_per_testcase(Func, Config) when is_atom(Func), is_list(Config) -> [{testcase, Func}|Config]. end_per_testcase(Func, Config) when is_atom(Func), is_list(Config) -> + %% Restore max_heap_size to default value. + erlang:system_flag(max_heap_size, + #{size => 0, + kill => true, + error_logger => true}), ok. fun_spawn(Fun) -> -- cgit v1.2.3 From ff1234e3eba1889a9b8cd81cc03f38cac78a67fe Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= Date: Thu, 1 Jun 2017 10:32:13 +0200 Subject: Robustify process_SUITE:spawn_opt_max_heap_size/1 process_SUITE starts os_mon in init_per_suite/1. Therefore, there may be occasional alarm info messages received. Make sure that they are ignored and don't cause the test fail to fail. Also, get rid of the flush/0 function that discards all messages in the message queue. Instead, be more selective and only discard {error, ...} messages. --- erts/emulator/test/process_SUITE.erl | 28 ++++++++++++++++++++-------- 1 file changed, 20 insertions(+), 8 deletions(-) diff --git a/erts/emulator/test/process_SUITE.erl b/erts/emulator/test/process_SUITE.erl index 180311202f..4204d12eb3 100644 --- a/erts/emulator/test/process_SUITE.erl +++ b/erts/emulator/test/process_SUITE.erl @@ -2074,6 +2074,7 @@ max_heap_size_test(Option, Size, Kill, ErrorLogger) -> end, if ErrorLogger -> receive + %% There must be at least one error message. {error, _, {emulator, _, [Pid|_]}} -> ok end; @@ -2086,22 +2087,33 @@ max_heap_size_test(Option, Size, Kill, ErrorLogger) -> {'DOWN', Ref, process, Pid, die} -> ok end, - flush(); + %% If the process was not killed, the limit may have + %% been reached more than once and there may be + %% more {error, ...} messages left. + receive_error_messages(Pid); true -> ok end, + + %% Make sure that there are no unexpected messages. + receive_unexpected(). + +receive_error_messages(Pid) -> receive - M -> - ct:fail({unexpected_message, M}) - after 10 -> + {error, _, {emulator, _, [Pid|_]}} -> + receive_error_messages(Pid) + after 1000 -> ok end. -flush() -> +receive_unexpected() -> receive - _M -> - flush() - after 1000 -> + {info_report, _, _} -> + %% May be an alarm message from os_mon. Ignore. + receive_unexpected(); + M -> + ct:fail({unexpected_message, M}) + after 10 -> ok end. -- cgit v1.2.3 From 6a18795542432fb3f8ba1d4306ec28d6a9250f3e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= Date: Thu, 1 Jun 2017 14:51:20 +0200 Subject: Remove obsolete old_scheduler_SUITE old_scheduler_SUITE was written before the SMP emulator. It does not test anything that scheduler_SUITE does not test. Keeping it wastes time when running the emulator test suite, and does not help us find any bugs that scheduler_SUITE would find. --- erts/emulator/test/Makefile | 1 - erts/emulator/test/old_scheduler_SUITE.erl | 384 ----------------------------- 2 files changed, 385 deletions(-) delete mode 100644 erts/emulator/test/old_scheduler_SUITE.erl diff --git a/erts/emulator/test/Makefile b/erts/emulator/test/Makefile index 2479ccc01f..fcd7244ae9 100644 --- a/erts/emulator/test/Makefile +++ b/erts/emulator/test/Makefile @@ -117,7 +117,6 @@ MODULES= \ tracer_SUITE \ tracer_test \ scheduler_SUITE \ - old_scheduler_SUITE \ port_trace_SUITE \ unique_SUITE \ z_SUITE \ diff --git a/erts/emulator/test/old_scheduler_SUITE.erl b/erts/emulator/test/old_scheduler_SUITE.erl deleted file mode 100644 index 8515a87df8..0000000000 --- a/erts/emulator/test/old_scheduler_SUITE.erl +++ /dev/null @@ -1,384 +0,0 @@ -%% -%% %CopyrightBegin% -%% -%% Copyright Ericsson AB 2004-2016. All Rights Reserved. -%% -%% Licensed under the Apache License, Version 2.0 (the "License"); -%% you may not use this file except in compliance with the License. -%% You may obtain a copy of the License at -%% -%% http://www.apache.org/licenses/LICENSE-2.0 -%% -%% Unless required by applicable law or agreed to in writing, software -%% distributed under the License is distributed on an "AS IS" BASIS, -%% WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -%% See the License for the specific language governing permissions and -%% limitations under the License. -%% -%% %CopyrightEnd% -%% - --module(old_scheduler_SUITE). - --include_lib("common_test/include/ct.hrl"). - --export([all/0, suite/0, - init_per_testcase/2, end_per_testcase/2]). --export([equal/1, many_low/1, few_low/1, max/1, high/1]). - -suite() -> - [{ct_hooks,[ts_install_cth]}, - {timetrap, {minutes, 11}}]. - -all() -> - case catch erlang:system_info(modified_timing_level) of - Level when is_integer(Level) -> - {skipped, - "Modified timing (level " ++ - integer_to_list(Level) ++ - ") is enabled. Testcases gets messed " - "up by modfied timing."}; - _ -> [equal, many_low, few_low, max, high] - end. - - -%%----------------------------------------------------------------------------------- -%% TEST SUITE DESCRIPTION -%% -%% The test case function spawns two controlling processes: Starter and Receiver. -%% Starter spawns a number of prio A and a number of prio B test processes. Each -%% test process loops for a number of times, sends a report to the Receiver, then -%% loops again. For each report, the Receiver increases a counter that corresponds -%% to the priority of the sender. After a certain amount of time, the Receiver -%% sends the collected data to the main test process and waits for the test case -%% to terminate. From this data, it's possible to calculate the average run time -%% relationship between the prio A and B test processes. -%% -%% Note that in order to be able to run tests with high or max prio test processes, -%% the main test process and the Receiver needs to run at max prio, or they will -%% be starved by the test processes. The controlling processes must not wait for -%% messages from a normal (or low) prio process while max or high prio test processes -%% are running (which happens e.g. if an io function is called). -%%----------------------------------------------------------------------------------- - -init_per_testcase(_Case, Config) -> - %% main test process needs max prio - Prio = process_flag(priority, max), - MS = erlang:system_flag(multi_scheduling, block_normal), - [{prio,Prio},{multi_scheduling, MS}|Config]. - -end_per_testcase(_Case, Config) -> - erlang:system_flag(multi_scheduling, unblock_normal), - Prio=proplists:get_value(prio, Config), - process_flag(priority, Prio), - ok. - -ok(Config) when is_list(Config) -> - case proplists:get_value(multi_scheduling, Config) of - blocked -> - {comment, - "Multi-scheduling blocked during test. This testcase was not " - "written to work with multiple schedulers."}; - _ -> ok - end. - -%% Run equal number of low and normal prio processes. - -equal(Config) when is_list(Config) -> - Self = self(), - - %% specify number of test processes to run - Normal = {normal,500}, - Low = {low,500}, - - %% specify time of test (in seconds) - Time = 30, - - %% start controllers - Receiver = - spawn(fun() -> receiver(erlang:monotonic_time(), Time, Self, Normal, Low) end), - Starter = - spawn(fun() -> starter(Normal, Low, Receiver) end), - - %% receive test data from Receiver - {NRs,NAvg,LRs,LAvg,Ratio} = - receive - {Receiver,Res} -> Res - end, - - %% stop controllers and test processes - exit(Starter, kill), - exit(Receiver, kill), - - io:format("Reports: ~w normal (~w/proc), ~w low (~w/proc). Ratio: ~w~n", - [NRs,NAvg,LRs,LAvg,Ratio]), - - %% runtime ratio between normal and low should be ~8 - if Ratio < 7.5 ; Ratio > 8.5 -> - ct:fail({bad_ratio,Ratio}); - true -> - ok(Config) - end. - - -%% Run many low and few normal prio processes. - -many_low(Config) when is_list(Config) -> - Self = self(), - Normal = {normal,1}, - Low = {low,1000}, - - %% specify time of test (in seconds) - Time = 30, - - Receiver = - spawn(fun() -> receiver(erlang:monotonic_time(), Time, Self, Normal, Low) end), - Starter = - spawn(fun() -> starter(Normal, Low, Receiver) end), - {NRs,NAvg,LRs,LAvg,Ratio} = - receive - {Receiver,Res} -> Res - end, - exit(Starter, kill), - exit(Receiver, kill), - io:format("Reports: ~w normal (~w/proc), ~w low (~w/proc). Ratio: ~w~n", - [NRs,NAvg,LRs,LAvg,Ratio]), - if Ratio < 7.5 ; Ratio > 8.5 -> - ct:fail({bad_ratio,Ratio}); - true -> - ok(Config) - end. - - -%% Run few low and many normal prio processes. - -few_low(Config) when is_list(Config) -> - Self = self(), - Normal = {normal,1000}, - Low = {low,1}, - - %% specify time of test (in seconds) - Time = 30, - - Receiver = - spawn(fun() -> receiver(erlang:monotonic_time(), Time, Self, Normal, Low) end), - Starter = - spawn(fun() -> starter(Normal, Low, Receiver) end), - {NRs,NAvg,LRs,LAvg,Ratio} = - receive - {Receiver,Res} -> Res - end, - exit(Starter, kill), - exit(Receiver, kill), - io:format("Reports: ~w normal (~w/proc), ~w low (~w/proc). Ratio: ~w~n", - [NRs,NAvg,LRs,LAvg,Ratio]), - if Ratio < 7.0 ; Ratio > 8.5 -> - ct:fail({bad_ratio,Ratio}); - true -> - ok(Config) - end. - - -%% Run max prio processes and verify they get at least as much -%% runtime as high, normal and low. - -max(Config) when is_list(Config) -> - max = process_flag(priority, max), % should already be max (init_per_tc) - Self = self(), - Max = {max,2}, - High = {high,2}, - Normal = {normal,100}, - Low = {low,100}, - - %% specify time of test (in seconds) - Time = 30, - - Receiver1 = - spawn(fun() -> receiver(erlang:monotonic_time(), Time, Self, Max, High) end), - Starter1 = - spawn(fun() -> starter(Max, High, Receiver1) end), - {M1Rs,M1Avg,HRs,HAvg,Ratio1} = - receive - {Receiver1,Res1} -> Res1 - end, - exit(Starter1, kill), - exit(Receiver1, kill), - io:format("Reports: ~w max (~w/proc), ~w high (~w/proc). Ratio: ~w~n", - [M1Rs,M1Avg,HRs,HAvg,Ratio1]), - if Ratio1 < 1.0 -> - ct:fail({bad_ratio,Ratio1}); - true -> - ok(Config) - end, - - Receiver2 = - spawn(fun() -> receiver(erlang:monotonic_time(), Time, Self, Max, Normal) end), - Starter2 = - spawn(fun() -> starter(Max, Normal, Receiver2) end), - {M2Rs,M2Avg,NRs,NAvg,Ratio2} = - receive - {Receiver2,Res2} -> Res2 - end, - exit(Starter2, kill), - exit(Receiver2, kill), - io:format("Reports: ~w max (~w/proc), ~w normal (~w/proc). Ratio: ~w~n", - [M2Rs,M2Avg,NRs,NAvg,Ratio2]), - if Ratio2 < 1.0 -> - ct:fail({bad_ratio,Ratio2}); - true -> - ok - end, - - Receiver3 = - spawn(fun() -> receiver(erlang:monotonic_time(), Time, Self, Max, Low) end), - Starter3 = - spawn(fun() -> starter(Max, Low, Receiver3) end), - {M3Rs,M3Avg,LRs,LAvg,Ratio3} = - receive - {Receiver3,Res3} -> Res3 - end, - exit(Starter3, kill), - exit(Receiver3, kill), - io:format("Reports: ~w max (~w/proc), ~w low (~w/proc). Ratio: ~w~n", - [M3Rs,M3Avg,LRs,LAvg,Ratio3]), - if Ratio3 < 1.0 -> - ct:fail({bad_ratio,Ratio3}); - true -> - ok(Config) - end. - - -%% Run high prio processes and verify they get at least as much -%% runtime as normal and low. - -high(Config) when is_list(Config) -> - max = process_flag(priority, max), % should already be max (init_per_tc) - Self = self(), - High = {high,2}, - Normal = {normal,100}, - Low = {low,100}, - - %% specify time of test (in seconds) - Time = 30, - - Receiver1 = - spawn(fun() -> receiver(erlang:monotonic_time(), Time, Self, High, Normal) end), - Starter1 = - spawn(fun() -> starter(High, Normal, Receiver1) end), - {H1Rs,H1Avg,NRs,NAvg,Ratio1} = - receive - {Receiver1,Res1} -> Res1 - end, - exit(Starter1, kill), - exit(Receiver1, kill), - io:format("Reports: ~w high (~w/proc), ~w normal (~w/proc). Ratio: ~w~n", - [H1Rs,H1Avg,NRs,NAvg,Ratio1]), - if Ratio1 < 1.0 -> - ct:fail({bad_ratio,Ratio1}); - true -> - ok - end, - - Receiver2 = - spawn(fun() -> receiver(erlang:monotonic_time(), Time, Self, High, Low) end), - Starter2 = - spawn(fun() -> starter(High, Low, Receiver2) end), - {H2Rs,H2Avg,LRs,LAvg,Ratio2} = - receive - {Receiver2,Res2} -> Res2 - end, - exit(Starter2, kill), - exit(Receiver2, kill), - io:format("Reports: ~w high (~w/proc), ~w low (~w/proc). Ratio: ~w~n", - [H2Rs,H2Avg,LRs,LAvg,Ratio2]), - if Ratio2 < 1.0 -> - ct:fail({bad_ratio,Ratio2}); - true -> - ok(Config) - end. - - -%%----------------------------------------------------------------------------------- -%% Controller processes and help functions -%%----------------------------------------------------------------------------------- - -receiver(T0, TimeSec, Main, {P1,P1N}, {P2,P2N}) -> - %% prio should be max so that mailbox doesn't overflow - process_flag(priority, max), - receiver(T0, TimeSec*1000, Main, P1,P1N,0, P2,P2N,0, 100000). - -%% uncomment lines below to get life sign (debug) -receiver(T0, Time, Main, P1,P1N,P1Rs, P2,P2N,P2Rs, 0) -> - % T = erlang:convert_time_unit(erlang:monotonic_time() - T0, native, millisecond), - % erlang:display({round(T/1000),P1Rs,P2Rs}), - receiver(T0, Time, Main, P1,P1N,P1Rs, P2,P2N,P2Rs, 100000); - -receiver(T0, Time, Main, P1,P1N,P1Rs, P2,P2N,P2Rs, C) -> - Remain = Time - erlang:convert_time_unit(erlang:monotonic_time() - T0, - native, millisecond), % test time remaining - Remain1 = if Remain < 0 -> - 0; - true -> - Remain - end, - {P1Rs1,P2Rs1} = - receive - {_Pid,P1} -> % report from a P1 process - {P1Rs+1,P2Rs}; - {_Pid,P2} -> % report from a P2 process - {P1Rs,P2Rs+1} - after Remain1 -> - {P1Rs,P2Rs} - end, - if Remain > 0 -> % keep going - receiver(T0, Time, Main, P1,P1N,P1Rs1, P2,P2N,P2Rs1, C-1); - true -> % finish - %% calculate results and send to main test process - P1Avg = P1Rs1/P1N, - P2Avg = P2Rs1/P2N, - Ratio = if P2Avg < 1.0 -> P1Avg; - true -> P1Avg/P2Avg - end, - Main ! {self(),{P1Rs1,round(P1Avg),P2Rs1,round(P2Avg),Ratio}}, - flush_loop() - end. - -starter({P1,P1N}, {P2,P2N}, Receiver) -> - %% start N1 processes with prio P1 - start_p(P1, P1N, Receiver), - %% start N2 processes with prio P2 - start_p(P2, P2N, Receiver), - erlang:display({started,P1N+P2N}), - flush_loop(). - -start_p(_, 0, _) -> - ok; -start_p(Prio, N, Receiver) -> - spawn_link(fun() -> p(Prio, Receiver) end), - start_p(Prio, N-1, Receiver). - -p(Prio, Receiver) -> - %% set process priority - process_flag(priority, Prio), - p_loop(0, Prio, Receiver). - -p_loop(100, Prio, Receiver) -> - receive after 0 -> ok end, - %% if Receiver gone, we're done - case is_process_alive(Receiver) of - false -> exit(bye); - true -> ok - end, - %% send report - Receiver ! {self(),Prio}, - p_loop(0, Prio, Receiver); - -p_loop(N, Prio, Receiver) -> - p_loop(N+1, Prio, Receiver). - - -flush_loop() -> - receive _ -> - ok - end, - flush_loop(). -- cgit v1.2.3 From 7aa347e1f911f1bcef026c08e32cc4eed280ce38 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= Date: Thu, 1 Jun 2017 15:16:00 +0200 Subject: same_time_yielding/1: Avoid failing if there are many schedulers On a computer with 32 schedulers, there would be 9632 (301*32) 'timeout' message in the receive queue. Receiving them with a selective receive (matching on the timer ID) is quite slow. Change the test case to read out the queue in the order the messages are stored in the queue. --- erts/emulator/test/timer_bif_SUITE.erl | 24 ++++++++++++++++++++---- 1 file changed, 20 insertions(+), 4 deletions(-) diff --git a/erts/emulator/test/timer_bif_SUITE.erl b/erts/emulator/test/timer_bif_SUITE.erl index 7cbd93a0f3..a977eb41c4 100644 --- a/erts/emulator/test/timer_bif_SUITE.erl +++ b/erts/emulator/test/timer_bif_SUITE.erl @@ -488,24 +488,40 @@ registered_process(Config) when is_list(Config) -> same_time_yielding(Config) when is_list(Config) -> Mem = mem(), + Ref = make_ref(), SchdlrsOnln = erlang:system_info(schedulers_online), Tmo = erlang:monotonic_time(millisecond) + 3000, Tmrs = lists:map(fun (I) -> process_flag(scheduler, (I rem SchdlrsOnln) + 1), - erlang:start_timer(Tmo, self(), hej, [{abs, true}]) + erlang:start_timer(Tmo, self(), Ref, [{abs, true}]) end, lists:seq(1, (?TIMEOUT_YIELD_LIMIT*3+1)*SchdlrsOnln)), true = mem_larger_than(Mem), - lists:foreach(fun (Tmr) -> receive {timeout, Tmr, hej} -> ok end end, Tmrs), + receive_all_timeouts(length(Tmrs), Ref), Done = erlang:monotonic_time(millisecond), true = Done >= Tmo, + MsAfterTmo = Done - Tmo, + io:format("Done ~p ms after Tmo\n", [MsAfterTmo]), case erlang:system_info(build_type) of - opt -> true = Done < Tmo + 200; - _ -> true = Done < Tmo + 1000 + opt -> + true = MsAfterTmo < 200; + _ -> + true = MsAfterTmo < 1000 end, Mem = mem(), ok. +%% Read out all timeouts in receive queue order. This is efficient +%% even if there are very many messages. + +receive_all_timeouts(0, _Ref) -> + ok; +receive_all_timeouts(N, Ref) -> + receive + {timeout, _Tmr, Ref} -> + receive_all_timeouts(N-1, Ref) + end. + same_time_yielding_with_cancel(Config) when is_list(Config) -> same_time_yielding_with_cancel_test(false, false). -- cgit v1.2.3 From 46b62dea54a230e46bbadfec7f119ff5ab513e6a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= Date: Fri, 2 Jun 2017 11:41:48 +0200 Subject: Stabilize call_with_huge_message_queue/1 Time measurements are always tricky, particulary on virtual hosts. On one particular virtual host, the measured times were often 0. Measuring the time for 100 calls instead of 10 calls helps, but 0 can still be returned, so we will also need to discard measurements that return 0 and try again. --- erts/emulator/test/receive_SUITE.erl | 55 +++++++++++++++++++++++------------- 1 file changed, 35 insertions(+), 20 deletions(-) diff --git a/erts/emulator/test/receive_SUITE.erl b/erts/emulator/test/receive_SUITE.erl index 83653a7a36..c7d5a3f5a0 100644 --- a/erts/emulator/test/receive_SUITE.erl +++ b/erts/emulator/test/receive_SUITE.erl @@ -39,25 +39,43 @@ groups() -> call_with_huge_message_queue(Config) when is_list(Config) -> Pid = spawn_link(fun echo_loop/0), - - {Time,ok} = tc(fun() -> calls(10, Pid) end), - - [self() ! {msg,N} || N <- lists:seq(1, 500000)], + _WarmUpTime = time_calls(Pid), + Time = time_calls(Pid), + _ = [self() ! {msg,N} || N <- lists:seq(1, 500000)], + io:format("Time for empty message queue: ~p", [Time]), erlang:garbage_collect(), - {NewTime1,ok} = tc(fun() -> calls(10, Pid) end), - {NewTime2,ok} = tc(fun() -> calls(10, Pid) end), + call_with_huge_message_queue_1(Pid, Time, 5). + +call_with_huge_message_queue_1(_Pid, _Time, 0) -> + ct:fail(bad_ratio); +call_with_huge_message_queue_1(Pid, Time, NumTries) -> + HugeTime = time_calls(Pid), + io:format("Time for huge message queue: ~p", [HugeTime]), + + case (HugeTime+1) / (Time+1) of + Q when Q < 10 -> + ok; + Q -> + io:format("Too high ratio: ~p\n", [Q]), + call_with_huge_message_queue_1(Pid, Time, NumTries-1) + end. - io:format("Time for empty message queue: ~p", [Time]), - io:format("Time1 for huge message queue: ~p", [NewTime1]), - io:format("Time2 for huge message queue: ~p", [NewTime2]), - - case hd(lists:sort([(NewTime1+1) / (Time+1), (NewTime2+1) / (Time+1)])) of - Q when Q < 10 -> - ok; - Q -> - ct:fail("Best Q = ~p", [Q]) - end, - ok. +%% Time a number calls. Try to avoid returning a zero time. +time_calls(Pid) -> + time_calls(Pid, 10). + +time_calls(_Pid, 0) -> + 0; +time_calls(Pid, NumTries) -> + case timer:tc(fun() -> calls(Pid) end) of + {0,ok} -> + time_calls(Pid, NumTries-1); + {Time,ok} -> + Time + end. + +calls(Pid) -> + calls(100, Pid). calls(0, _) -> ok; calls(N, Pid) -> @@ -108,6 +126,3 @@ echo_loop() -> Pid ! {Ref,Msg}, echo_loop() end. - -tc(Fun) -> - timer:tc(erlang, apply, [Fun,[]]). -- cgit v1.2.3