diff options
Diffstat (limited to 'erts/doc/src/notes.xml')
| -rw-r--r-- | erts/doc/src/notes.xml | 757 |
1 files changed, 757 insertions, 0 deletions
diff --git a/erts/doc/src/notes.xml b/erts/doc/src/notes.xml index 4cef9669dd..6b6a3374d1 100644 --- a/erts/doc/src/notes.xml +++ b/erts/doc/src/notes.xml @@ -30,6 +30,763 @@ </header> <p>This document describes the changes made to the ERTS application.</p> +<section><title>Erts 5.9.1.1</title> + + <section><title>Fixed Bugs and Malfunctions</title> + <list> + <item> + <p> + If threads support for the runtime system had been + disabled at compile time (<c>--disable-threads</c> had + been passed to <c>configure</c>), and the <c>+A</c> + command line argument of <c>erl</c> was passed when + starting the runtime system, <seealso + marker="erl_driver#driver_system_info">driver_system_info()</seealso> + erroneously claimed that the runtime system had async + threads even though it had not.</p> + <p> + Due to this bug the file driver did not split tasks into + smaller chunks, but instead completed the whole task at + once, i.e., the scheduler got occupied with I/O for a + longer time than intended.</p> + <p> + Own Id: OTP-10059</p> + </item> + </list> + </section> + + + <section><title>Improvements and New Features</title> + <list> + <item> + <p> + A proposal for a new scheduler wakeup strategy has been + implemented. For more information see the documentation + of the <seealso marker="erl#+sws">+sws</seealso> command + line argument of <c>erl</c>.</p> + <p> + Own Id: OTP-10033 Aux Id: Seq12025 </p> + </item> + <item> + <p> + A switch for configuration of busy wait length for + scheduler threads has been added. For more information + see the documentation of the <seealso + marker="erl#+sbwt">+sbwt</seealso> command line argument + of <c>erl</c>.</p> + <p> + Own Id: OTP-10044 Aux Id: Seq11976 </p> + </item> + <item> + <p> + The extra memory barriers introduced by bug-fix OTP-9281 + were unnecessarily used also on tables without the + <c>write_concurrency</c> option enabled. This could + unnecessarily degrade performance of ETS tables without + <c>write_concurrency</c> on some hardware (e.g. PowerPC) + while not effecting performance at all on other hardware + (e.g. x86/x86_64).</p> + <p> + OTP-9281 (R14B03): ETS tables using the + <c>write_concurrency</c> option could potentially get + into an internally inconsistent state.</p> + <p> + Own Id: OTP-10048 Aux Id: OTP-9281 </p> + </item> + </list> + </section> + +</section> + +<section><title>Erts 5.9.1</title> + + <section><title>Fixed Bugs and Malfunctions</title> + <list> + <item> + <p><c>erlang:system_profile</c> errorneous profiled the + profiler process when observing runnable processes. This + has been corrected. </p> + <p> + Own Id: OTP-9849</p> + </item> + <item> + <p>Calling trace_info/2 asking for information about a + function that had native could could crash the run-time + system.</p> + <p> + Own Id: OTP-9886</p> + </item> + <item> + <p> + reduce smp locking time range in erts_garbage_collect + (thanks to Jovi Zhang)</p> + <p> + Own Id: OTP-9912</p> + </item> + <item> + <p> + Fix typo in supervisor behaviour doc (Thanks to Ricardo + Catalinas Jim�nez)</p> + <p> + Own Id: OTP-9924</p> + </item> + <item> + <p> + Correct spelling of registered (Thanks to Richard + Carlsson)</p> + <p> + Own Id: OTP-9925</p> + </item> + <item> + <p> + erts: Remove unused variable (Thanks to Jovi Zhang)</p> + <p> + Own Id: OTP-9926</p> + </item> + <item> + <p> + Fix bug in ETS with <c>compressed</c> option and + insertion of term containing large integers (>2G) on + 64-bit machines. Seen to cause emulator crash. (Thanks to + Diego Llarrull for excellent bug report)</p> + <p> + Own Id: OTP-9932</p> + </item> + <item> + <p> + Handle Linux OS where /sys/devices/system/node is only + readable by root. Fallback to /sys/devices/system/cpu for + topology info.</p> + <p> + Own Id: OTP-9978</p> + </item> + <item> + <p> When an escript ends now all printout to standard + output and standard error gets out on the terminal. This + bug has been corrected by changing the behaviour of + erlang:halt/0,1, which should fix the same problem for + other escript-like applications, i.e that data stored in + the output port driver buffers got lost when printing on + a TTY and exiting through erlang:halt/0,1. </p> + <p> The BIF:s erlang:halt/0,1 has gotten improved + semantics and there is a new BIF erlang:halt/2 to + accomplish something like the old semantics. See the + documentation. </p> + <p> Now erlang:halt/0 and erlang:halt/1 with an integer + argument will close all ports and allow all pending async + threads operations to finish before exiting the emulator. + Previously erlang:halt/0 and erlang:halt(0) would just + wait for pending async threads operations but not close + ports. And erlang:halt/1 with a non-zero integer argument + would not even wait for pending async threads operations. + </p> + <p> To roughly the old behaviour, to not wait for ports + and async threads operations when you exit the emulator, + you use erlang:halt/2 with an integer first argument and + an option list containing {flush,false} as the second + argument. Note that now is flushing not dependant of the + exit code, and you can not only flush async threads + operations which we deemed as a strange behaviour anyway. + </p> + <p>Also, erlang:halt/1,2 has gotten a new feature: If the + first argument is the atom 'abort' the emulator is + aborted producing a core dump, if the operating system so + allows. </p> + <p> + Own Id: OTP-9985</p> + </item> + <item> + <p> + Added check to inet driver to avoid building on operating + systems that do not yet have IPv6 compatible socket API. + (Thanks to Peer Stritzinger)</p> + <p> + Own Id: OTP-9996</p> + </item> + <item> + <p> + Fix bug when the number of CPUs actually found is lower + than the configured value. (Thanks to Benjamin + Herrenschmidt)</p> + <p> + Own Id: OTP-10004</p> + </item> + <item> + <p> + The runtime system without SMP support and without thread + support erroneously busy waited when no work was present. + This bug first appeared in <c>erts-5.9</c>.</p> + <p> + Own Id: OTP-10019</p> + </item> + <item> + <p> + Various typographical errors corrected in documentation + for common_test, driver, erl_driver and windows + installation instructions. (Thanks to Tuncer Ayaz)</p> + <p> + Own Id: OTP-10037</p> + </item> + <item> + <p> + Fix memory leak caused by race on exiting process</p> + <p> + Own Id: OTP-10041</p> + </item> + </list> + </section> + + + <section><title>Improvements and New Features</title> + <list> + <item> + <p>Add <c>erlang:statistics(scheduler_wall_time)</c> to + ensure correct determination of scheduler utilization. + Measuring scheduler utilization is strongly preferred + over CPU utilization, since CPU utilization gives very + poor indications of actual scheduler/vm usage.</p> + <p> + Own Id: OTP-9858</p> + </item> + <item> + <p> + ERTS internal API improvements. In some cases the amount + of atomic read operations needed have been reduced due to + this.</p> + <p> + Own Id: OTP-9922</p> + </item> + <item> + <p> + The DTrace source patch from Scott Lystig Fritchie is + integrated in the source tree. Using an emulator with + dtrace probe is still not supported for production use, + but may be a valuable debugging tool. Configure with + --with-dynamic-trace=dtrace (or + --with-dynamic-trace=systemtap) to create a build with + dtrace probes enabled. See runtime_tools for + documentation and examples.</p> + <p> + Own Id: OTP-10017</p> + </item> + </list> + </section> + + + <section><title>Known Bugs and Problems</title> + <list> + <item> + <p> + enif_make_copy may invalidate enif_inspect_binary.</p> + <p> + Own Id: OTP-9828</p> + </item> + </list> + </section> + +</section> + +<section><title>Erts 5.9.0.1</title> + + <section><title>Fixed Bugs and Malfunctions</title> + <list> + <item> + <p> + A feature test for the <c>lwsync</c> instruction + performed on PowerPC hardware at runtime system startup + got into an eternal loop if the instruction was not + supported. This bug was introduced in erts-5.9/OTP-R15B.</p> + <p> + Own Id: OTP-9843</p> + </item> + <item> + <p> + I/O events could potentially be delayed for ever when + enabling kernel-poll on a non-SMP runtime system + executing on Solaris. When also combined with + async-threads the runtime system hung before completing + the boot phase. This bug was introduced in + erts-5.9/OTP-R15B.</p> + <p> + Own Id: OTP-9844</p> + </item> + </list> + </section> + +</section> + +<section><title>Erts 5.9</title> + + <section><title>Fixed Bugs and Malfunctions</title> + <list> + <item> + <p> + Honor option <c>packet_size</c> for http packet parsing + by both TCP socket and <c>erlang:decode_packet</c>. This + gives the ability to accept HTTP headers larger than the + default setting, but also avoid DoS attacks by accepting + lines only up to whatever length you wish to allow. For + consistency, packet type <c>line</c> also honor option + <c>packet_size</c>. (Thanks to Steve Vinoski)</p> + <p> + Own Id: OTP-9389</p> + </item> + <item> + <p> A few contracts in the <c>lists</c> module have been + corrected. </p> + <p> + Own Id: OTP-9616</p> + </item> + <item> + <p> + The Unicode noncharacter code points 16#FFFE and 16#FFFE + were not allowed to be encoded or decoded using the + <c>unicode</c> module or bit syntax. That was + inconsistent with the other noncharacters 16#FDD0 to + 16#FDEF that could be encoded/decoded. To resolve the + inconsistency, 16#FFFE and 16#FFFE can now be encoded and + decoded. (Thanks to Alisdair Sullivan.)</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-9624</p> + </item> + <item> + <p> + Make epp search directory of current file first when + including another file This completes a partial fix in + R11 that only worked for include_lib(). (Thanks to + Richard Carlsson)</p> + <p> + Own Id: OTP-9645</p> + </item> + <item> + <p> + Fixed memory leak in + <c>enif_inspect_io_list_as_binary</c> when applied on a + process independent environment.</p> + <p> + Own Id: OTP-9668</p> + </item> + <item> + <p>The number of beam catches allowed in code are no + longer statically defined and will grow according to its + need.</p> + <p> + Own Id: OTP-9692</p> + </item> + <item> + <p> + Add missing parenthesis in heart doc.</p> + <p> + Add missing spaces in the Reference Manual distributed + section.</p> + <p> + In the HTML version of the doc those spaces are necessary + to separate those words.</p> + <p> + Own Id: OTP-9693</p> + </item> + <item> + <p> + Fixes module erlang doc style: option description (Thanks + to Ricardo Catalinas Jim�nez)</p> + <p> + Own Id: OTP-9697</p> + </item> + <item> + <p> + Specifying a scope to binary:match/3 when using multiple + searchstrings resulted in faulty return values. This is + now corrected.</p> + <p> + Own Id: OTP-9701</p> + </item> + <item> + <p> + The runtime system crashed if more than one thread tried + to exit the runtime system at the same time.</p> + <p> + Own Id: OTP-9705</p> + </item> + <item> + <p> + Fix documentation for erlang:process_flag/2</p> + <p> + For the subsection about process_flag(save_calls, N) + there's an unrelated paragraph about process priorities + which was copied from the preceeding subsection regarding + process_flag(priority, Level). (Thanks to Filipe David + Manana)</p> + <p> + Own Id: OTP-9714</p> + </item> + <item> + <p> + Calls to <c>erlang:system_flag(schedulers_online, N)</c> + and/or <c>erlang:system_flag(multi_scheduling, + block|unblock)</c> could cause internal data used by this + functionality to get into an inconsistent state. When + this happened various problems occurred. This bug was + quite hard to trigger, so hopefully no-one has been + effected by it.</p> + <p> + A spinlock used by the run-queue management sometimes got + heavily contended. This code has now been rewritten, and + the spinlock has been removed.</p> + <p> + Own Id: OTP-9727</p> + </item> + <item> + <p> + Use libdlpi to get physical address (Thanks to Trond + Norbye)</p> + <p> + Own Id: OTP-9818</p> + </item> + </list> + </section> + + + <section><title>Improvements and New Features</title> + <list> + <item> + <p> An option list argument can now be passed to + <c>file:read_file_info/2, file:read_link_info/2</c> and + <c>file:write_file_info/3</c> and set time type + information in the call. Valid options are <c>{time, + local}, {time, universal}</c> and <c>{time, posix}</c>. + In the case of <c>posix</c> time no conversions are made + which makes the operation a bit faster. </p> + <p> + Own Id: OTP-7687</p> + </item> + <item> + <p>A number of memory allocation optimizations have been + implemented. Most optimizations reduce contention caused + by synchronization between threads during allocation and + deallocation of memory. Most notably:</p> <list> <item> + Synchronization of memory management in scheduler + specific allocator instances has been rewritten to use + lock-free synchronization. </item> <item> Synchronization + of memory management in scheduler specific pre-allocators + has been rewritten to use lock-free synchronization. + </item> <item> The 'mseg_alloc' memory segment allocator + now use scheduler specific instances instead of one + instance. Apart from reducing contention this also + ensures that memory allocators always create memory + segments on the local NUMA node on a NUMA system. </item> + </list> + <p> + Own Id: OTP-7775</p> + </item> + <item> + <p> + The ethread atomic memory operations API used by the + runtime system has been extended and improved.</p> + <p> + The ethread library now also performs runtime tests for + presence of hardware features, such as for example SSE2 + instructions, instead of requiring this to be determined + at compile time.</p> + <p> + All uses of the old deprecated atomic API in the runtime + system have been replaced with the use of the new atomic + API. In a lot of places this change imply a relaxation of + memory barriers used.</p> + <p> + Own Id: OTP-9014</p> + </item> + <item> + <p>gen_sctp:open/0-2 may now return + {error,eprotonosupport} if SCTP is not supported</p> + <p>gen_sctp:peeloff/1 has been implemented and creates a + one-to-one socket which also are supported now</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-9239</p> + </item> + <item> + <p> + Sendfile has been added to the file module's API. + sendfile/2 is used to read data from a file and send it + to a tcp socket using a zero copying mechanism if + available on that OS.</p> + <p> + Thanks to Tuncer Ayaz and Steve Vinovski for original + implementation</p> + <p> + Own Id: OTP-9240</p> + </item> + <item> + <p> + enif_get_reverse_list function added to nif API. This + function should be used to reverse small lists which are + deep within other structures making it impractical to do + the reverse in Erlang.</p> + <p> + Own Id: OTP-9392</p> + </item> + <item> + <p> + The deprecated concat_binary/1 BIF has been removed. Use + <c>list_to_binary</c> or <c>iolist_to_binary/1</c> + instead.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-9421</p> + </item> + <item> + <p>Erlang/OTP can now be built using parallel make if you + limit the number of jobs, for instance using '<c>make + -j6</c>' or '<c>make -j10</c>'. '<c>make -j</c>' does not + work at the moment because of some missing + dependencies.</p> + <p> + Own Id: OTP-9451</p> + </item> + <item> + <p>Line number and filename information are now included + in exception backtraces as a fourth element in the MFA + tuple. The information will be pretty-printed by the + shell and used by <c>common_test</c> to provide better + indication of where a test case.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-9468</p> + </item> + <item> + <p>All binary constants used to be handled as heap + binaries (i.e. the entire binary would be copied when + sent to another process). Binary constants larger than 64 + bytes are now refc binaries (i.e. the actual data in the + binary will not be copied when sent to another + process).</p> + <p> + Own Id: OTP-9486</p> + </item> + <item> + <p> + If a float and an integer is compared, the integer is + only converted to a float if the float datatype can + contain it. Otherwise the float is converted to an + integer.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-9497</p> + </item> + <item> + <p> + Add NIF function enif_is_number</p> + <p> + This function allows for easily determining if a term + represents or not a number (integer, float, small or + big).(Thanks to Filipe David Manana)</p> + <p> + Own Id: OTP-9629</p> + </item> + <item> + <p> + The ERTS internal system block functionality has been + replaced by new functionality for blocking the system. + The old system block functionality had contention issues + and complexity issues. The new functionality piggy-backs + on thread progress tracking functionality needed by newly + introduced lock-free synchronization in the runtime + system. When the functionality for blocking the system + isn't used, there is more or less no overhead at all. + This since the functionality for tracking thread progress + is there and needed anyway.</p> + <p> + Own Id: OTP-9631</p> + </item> + <item> + <p> + An ERTS internal, generic, many to one, lock-free queue + for communication between threads has been introduced. + The many to one scenario is very common in ERTS, so it + can be used in a lot of places in the future. Currently + it is used by scheduling of certain jobs, and the async + thread pool, but more uses are planned for the future.</p> + <p> + Drivers using the driver_async functionality are not + automatically locked to the system anymore, and can be + unloaded as any dynamically linked in driver.</p> + <p> + Scheduling of ready async jobs is now also interleaved in + between other jobs. Previously all ready async jobs were + performed at once.</p> + <p> + Own Id: OTP-9632</p> + </item> + <item> + <p> + Tuple funs (a two-element tuple with a module name and a + function) are now officially deprecated and will be + removed in R16. Use '<c>fun M:F/A</c>' instead. To make + you aware that your system uses tuple funs, the very + first time a tuple fun is applied, a warning will be sent + to the error logger.</p> + <p> + Own Id: OTP-9649</p> + </item> + <item> + <p> + Changed the internal BIF calling convention. Will make + simpler faster code and allow BIFs with an arbitrary + arity.</p> + <p> + Own Id: OTP-9662</p> + </item> + <item> + <p> + Windows native critical sections are now used internally + in the runtime system on Windows as mutex implementation. + This since they perform better under extreme contention + than our own implementation.</p> + <p> + Own Id: OTP-9671</p> + </item> + <item> + <p> + Convert some erl_nif macros into inline functions. Allow + for better compile time type checking. (Thanks to Tuncer + Ayaz)</p> + <p> + Own Id: OTP-9675</p> + </item> + <item> + <p> + The <c>+scl</c> command line flag has been added. It can + be used for disabling compaction of scheduler load. For + more information see the <c>erl(1)</c> documentation.</p> + <p> + Own Id: OTP-9695</p> + </item> + <item> + <p>The build system has been updated so that Erlang/OTP + can be built on Mac OS X Lion systems without a GCC + compiler. The INSTALL guide has been updated with + instructions on how to install a GCC compiler and build + Erlang/OTP with it, in order to get a run-time system + with better performance.</p> + <p> + Own Id: OTP-9712</p> + </item> + <item> + <p> + When loading a module, the system use to run on a single + scheduler during the entire loading process. This has + been changed to only take down the system just before + inserting the loaded code into the system tables, + resulting in a much shorter disruption if a module is + loaded in a busy system. (Suggested by Bob Ippolito.)</p> + <p> + Own Id: OTP-9720</p> + </item> + <item> + <p> + Possible to run HiPE without floating point exceptions + (FPE). Useful on platforms that lack reliable FPE. Slower + float operations compared to HiPE with FPE.</p> + <p> + Own Id: OTP-9724</p> + </item> + <item> + <p> + As of ERTS version 5.9 (OTP-R15B) the runtime system will + by default <em>not</em> bind schedulers to logical + processors.</p> + <p> + If the Erlang runtime system is the only operating system + process that binds threads to logical processors, this + improves the performance of the runtime system. However, + if other operating system processes (as for example + another Erlang runtime system) also bind threads to + logical processors, there might be a performance penalty + instead. In some cases this performance penalty might be + severe. Due to this, we change the default so that the + user must make an active decision in order to bind + schedulers.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-9726</p> + </item> + <item> + <p> + The use of <c>erlang:system_flag(scheduler_bind_type, + _)</c> and <c>erlang:system_flag(cpu_topology, _)</c> + have been deprecated and scheduled for removal in + erts-5.10/OTP-R16. For more information see the + documentation of <c>erlang:system_flag/2</c>.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-9749</p> + </item> + <item> + <p> + An ancient workaround for a Windows bug was removed from + the open_port code, open_port({spawn,...}...) is now + faster. Thanks to Daniel Goertzen.</p> + <p> + Own Id: OTP-9766</p> + </item> + <item> + <p> + The use of deprecated 32bit time_t on 32bit Windows is + removed.</p> + <p> + Own Id: OTP-9767</p> + </item> + <item> + <p> + The NIF <c>reload</c> mechanism is deprecated. Do not use + it as an upgrade method for live production systems. It + might be removed in future releases. It can still serve + as a development feature but a warning message will be + logged each time it is used.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-9771</p> + </item> + <item> + <p>The driver interface has been changed to enable 64-bit + aware drivers. Most importantly the return types for + ErlDrvEntry callbacks 'call' and 'control' has ben + enlarged which require drivers to be changed or they will + cause emulator crashes. See <seealso + marker="erl_driver#rewrites_for_64_bits"> Rewrites for + 64-bit driver interface </seealso> in the driver manual. + </p> + <p>Due to this driver <seealso + marker="erl_driver#version_management">version + management</seealso> is now mandatory. A driver that is + not written with version management or a driver that was + compiled with the wrong major version will be not be + loaded by the emulator.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-9795</p> + </item> + <item> + <p> + Eliminate use of deprecated regexp module</p> + <p> + Own Id: OTP-9810</p> + </item> + </list> + </section> + +</section> + <section><title>Erts 5.8.5</title> <section><title>Fixed Bugs and Malfunctions</title> |