diff options
Diffstat (limited to 'erts/doc')
| -rw-r--r-- | erts/doc/src/notes.xml | 519 |
1 files changed, 519 insertions, 0 deletions
diff --git a/erts/doc/src/notes.xml b/erts/doc/src/notes.xml index e70e314abe..42298e4824 100644 --- a/erts/doc/src/notes.xml +++ b/erts/doc/src/notes.xml @@ -30,6 +30,525 @@ </header> <p>This document describes the changes made to the ERTS application.</p> +<section><title>Erts 5.10</title> + + <section><title>Fixed Bugs and Malfunctions</title> + <list> + <item> + <p> + Set new peeled off SCTP socket to nonblocking socket + (Thanks to Jonas Falkevik)</p> + <p> + Own Id: OTP-10491</p> + </item> + <item> + <p> + Fix various typos (thanks to Tuncer Ayaz)</p> + <p> + Own Id: OTP-10611</p> + </item> + <item> + <p> + Fix fd leak when using async thread pool</p> + <p> When using the async thread pool, if an erlang + process asks to open a file and it gets shutdown/killed + while the file:open/2 call hasn't returned, it's possible + to leak a file descriptor against the target file. This + has now been fixed. (Thanks to Filipe David Manana)</p> + <p> + Own Id: OTP-10677</p> + </item> + <item> + <p> + Use sys/types.h instead of string.h to pull ssize_t + definition to erl_driver.h. This fixes build issue on + NetBSD. (Thanks to Yamamoto Takashi).</p> + <p> + Own Id: OTP-10699</p> + </item> + <item> + <p> + Arguments given with the -run or -s flags to erl are now + translated according to the file name encoding mode of + the runtime system.</p> + <p> + Own Id: OTP-10702</p> + </item> + <item> + <p> + The octet counters in the gen_tcp/inet interface could + behave in unexpected ways on 64bit platforms. The + behaviour is now as expected.</p> + <p> + Own Id: OTP-10746</p> + </item> + <item> + <p> + Certain linux kernels, most notably in redhat and CentOS + distribution, had a bug in writev which generated an + infinite loop in the tcp code of the VM. The bug is now + worked around.</p> + <p> + Own Id: OTP-10747</p> + </item> + <item> + <p> + A process that got killed (got an exit signal) while + operating on a compresseed file, could cause a + segmentation fault in the VM. This is now corrected. + Thanks to Filipe David Manana for identifying the problem + and submitting a solution.</p> + <p> + Own Id: OTP-10748</p> + </item> + <item> + <p> + Windows previously used three digit exponent in + formatting which caused difference between platforms, as + can be seen by float_to_list/1. This has now been fixed.</p> + <p> + Own Id: OTP-10751</p> + </item> + </list> + </section> + + + <section><title>Improvements and New Features</title> + <list> + <item> + <p> + A boolean socket option 'ipv6_v6only' for IPv6 sockets + has been added. The default value of the option is OS + dependent, so applications aiming to be portable should + consider using <c>{ipv6_v6only,true}</c> when creating an + <c>inet6</c> listening/destination socket, and if + neccesary also create an <c>inet</c> socket on the same + port for IPv4 traffic. See the documentation.</p> + <p> + Own Id: OTP-8928 Aux Id: kunagi-193 [104] </p> + </item> + <item> + <p>It is now allowed to define stubs for BIFs, to allow + type specs to be written for BIFs. For example, if there + is BIF called <c>lists:member/2</c>, a dummy definition + of <c>lists:member/2</c> is now allowed.</p> + <p> + Own Id: OTP-9861</p> + </item> + <item> + <p>Process optimizations. The most notable:</p> <list> + <item>New internal process table implementation allowing + for both parallel reads as well as writes. Especially + read operations have become really cheap. This reduce + contention in various situations. For example when, + spawning processes, terminating processes, sending + messages, etc.</item> <item>Optimizations of run queue + management reducing contention.</item> + <item>Optimizations of process state changes reducing + contention.</item> </list> <p>These changes imply changes + of the characteristics the system. Most notable: changed + timing in the system.</p> + <p> + Own Id: OTP-9892 Aux Id: OTP-10167 </p> + </item> + <item> + <p> + Non-blocking code loading. Earlier when an Erlang module + was loaded, all other execution in the VM were halted + while the load operation was carried out in single + threaded mode. Now modules are loaded without blocking + the VM. Processes may continue executing undisturbed in + parallel during the entire load operation. The load + operation is completed by making the loaded code visible + to all processes in a consistent way with one single + atomic instruction. Non-blocking code loading will + improve realtime characteristics when modules are + loaded/upgraded on a running SMP system.</p> + <p> + Own Id: OTP-9974</p> + </item> + <item> + <p>In the SMP emulator, turning on and off tracing will + no longer take down the system to single-scheduling. </p> + <p> + Own Id: OTP-10122</p> + </item> + <item> + <p>Remove VxWorks support</p> + <p> + Own Id: OTP-10146</p> + </item> + <item> + <p> + Added a general framework for executing benchmarks of + Erlang/OTP. Benchmarks for the Erlang VM and mnesia have + been incorporated in the framework. </p> + <p> + For details about how to add more benchmarks see + $ERL_TOP/HOWTO/BENCHMARKS.md in the source distribution.</p> + <p> + Own Id: OTP-10156</p> + </item> + <item> + <p> + Optimized deletion of ETS-tables which significantly + improves performance when large amounts of temporary + tables are used.</p> + <p> + This change imply changes of the characteristics the + system. Most notable: changed timing in the system.</p> + <p> + Own Id: OTP-10167 Aux Id: OTP-9892 </p> + </item> + <item> + <p> + Tuple funs (deprecated in R15B) are no longer supported.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-10170</p> + </item> + <item> + <p> + New internal header scheme for allocators</p> + <p> + Impact: Reduces size on object allocated in multiblock + carriers by one word</p> + <p> + Own Id: OTP-10273 Aux Id: kunagi-20 [20] </p> + </item> + <item> + <p>Major port improvements. The most notable:</p> <list> + <item>New internal port table implementation allowing for + both parallel reads as well as writes. Especially read + operations have become really cheap.This reduce + contention in various situations. For example when, + creating ports, terminating ports, etc. </item> + <item>Dynamic allocation of port structures. This allow + for a much larger maximum amount of ports allowed as a + default. The previous default of 1024 has been raised to + 65536. Maximum amount of ports can be set using the + <seealso marker="erts:erl#+Q">+Q</seealso> command line + flag of <seealso marker="erts:erl">erl(1)</seealso>. The + previously used environment variable <c>ERL_MAX_PORTS</c> + has been deprecated and scheduled for removal in + OTP-R17.</item> <item>Major rewrite of scheduling of port + tasks. Major benefits of the rewrite are reduced + contention on run queue locks, and reduced amount of + memory allocation operations needed. The rewrite was also + necessary in order to make it possible to schedule + signals from processes to ports.</item> <item>Improved + internal thread progress functionality for easy + management of unmanaged threads. This improvement was + necessary for the rewrite of the port task + scheduling.</item> <item>Rewrite of all process to port + signal implementations in order to make it possible to + schedule those operations. All port operations can now be + scheduled which allows for reduced lock contention on the + port lock as well as truly asynchronous communication + with ports.</item> <item>Optimized lookup of port handles + from drivers.</item> <item>Optimized driver lookup when + creating ports.</item> <item>Preemptable <seealso + marker="erts:erlang#ports-0">erlang:ports/0</seealso> + BIF.</item> <item>Improving responsiveness by bumping + reductions for a process calling a driver callback + directly.</item> </list> + <p>These changes imply changes of the characteristics of + the system. The most notable:</p> <taglist> <tag>Order of + signal delivery</tag> <item>The previous implementation + of the VM has delivered signals from processes to ports + in a synchronous stricter fashion than required by the + language. As of ERTS version 5.10, signals are truly + asynchronously delivered. The order of signal delivery + still adheres to the requirements of the language, but + only to the requirements. That is, some signal sequences + that previously always were delivered in one specific + order may now from time to time be delivered in different + orders. This may cause Erlang programs that have made + <em>false assumptions</em> about signal delivery order to + fail even though they previously succeeded. For more + information about signal ordering guarantees, see the + chapter on <seealso + marker="erts:communication">communication</seealso> in + the ERTS user's guide. The <seealso + marker="erts:erl#+n">+n</seealso> command line flag of + <seealso marker="erts:erl">erl(1)</seealso> can be + helpful when trying to find signaling order bugs in + Erlang code that have been exposed by these + changes.</item> <tag>Latency of signals sent from + processes to ports</tag> <item>Signals from processes to + ports where previously always delivered immediately. This + kept latency for such communication to a minimum, but it + could cause lock contention which was very expensive for + the system as a whole. In order to keep this latency low + also in the future, most signals from processes to ports + are by default still delivered immediately as long as no + conflicts occur. Such conflicts include not being able to + acquire the port lock, but also include other conflicts. + When a conflict occur, the signal will be scheduled for + delivery at a later time. A scheduled signal delivery may + cause a higher latency for this specific communication, + but improves the overall performance of the system since + it reduce lock contention between schedulers. The default + behavior of only scheduling delivery of these signals on + conflict can be changed by passing the <seealso + marker="erts:erl#+spp">+spp</seealso> command line flag + to <seealso marker="erts:erl">erl(1)</seealso>. The + behavior can also be changed on port basis using the + <seealso + marker="erts:erlang#open_port_parallelism">parallelism</seealso> + option of the <seealso + marker="erts:erlang#open_port-2">open_port/2</seealso> + BIF.</item> <tag>Execution time of the + <c>erlang:ports/0</c> BIF</tag> <item>Since <seealso + marker="erts:erlang#ports-0">erlang:ports/0</seealso> now + can be preempted, the responsiveness of the system as a + whole has been improved. A call to <c>erlang:ports/0</c> + may, however, take a much longer time to complete than + before. How much longer time heavily depends on the + system load.</item> <tag>Reduction cost of calling driver + callbacks</tag> <item>Calling a driver callback is quite + costly. This was previously not reflected in reduction + cost at all. Since the reduction cost now has increased, + a process performing lots of direct driver calls will be + scheduled out more frequently than before.</item> + </taglist> + <p><em>Potential incompatibilities</em>:</p> <list> + <item><c>driver_send_term()</c> has been deprecated and + has been scheduled for removal in OTP-R17. Replace usage + of <c>driver_send_term()</c> with usage of <seealso + marker="erts:erl_driver#erl_drv_send_term">erl_drv_send_term()</seealso>.</item> + <item><c>driver_output_term()</c> has been deprecated and + has been scheduled for removal in OTP-R17. Replace usage + of <c>driver_output_term()</c> with usage of <seealso + marker="erts:erl_driver#erl_drv_output_term">erl_drv_output_term()</seealso>.</item> + <item>The new function <seealso + marker="erts:erl_driver#erl_drv_busy_msgq_limits">erl_drv_busy_msgq_limits()</seealso> + has been added in order to able to control management of + port queues.</item> </list> + <p>The <seealso + marker="erts:erl_driver#version_management">driver API + version</seealso> has been bumped to 2.1 from 2.0 due to + the above changes in the driver API.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-10336 Aux Id: OTP-9892 </p> + </item> + <item> + <p> + The experimental support for packages has been removed.</p> + <p> + Own Id: OTP-10348 Aux Id: kunagi-316 [227] </p> + </item> + <item> + <p> + Wrong parameters when setting seq_trace-tokens from + within a trace-pattern could crash the VM. This is now + corrected.</p> + <p> + Own Id: OTP-10522</p> + </item> + <item> + <p> + Erlang specification 4.7.3 defines max tuple size to + 65535 elements It is now enforced to no more than + 16777215 elements (arity 24 bits)</p> + <p> + Previous edge cases (28 bits) were not validated and + could cause undefined behaviour.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-10633</p> + </item> + <item> + <p> + Add insert_element/3 and delete_element/2</p> + <p> + Own Id: OTP-10643</p> + </item> + <item> + <p> + The previous default of a maximum of 32768 simultaneous + processes has been raised to 262144. This value can be + changed using the the <seealso + marker="erl#+P">+P</seealso> command line flag of + <seealso marker="erl">erl(1)</seealso>. Note that the + value passed now is considered as a hint, and that actual + value chosen in most cases will be a power of two.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-10647 Aux Id: OTP-9892, OTP-10336 </p> + </item> + <item> + <p> + The previously (in R15) proposed scheduler wakeup + strategy is now used by default. This strategy is not as + quick to forget about previous overload as the previous + strategy.</p> + <p> + This change imply changes of the characteristics the + system. Most notable: When a small overload comes and + then disappears repeatedly, the system will for a bit + longer time be willing to wake up schedulers than before. + Timing in the system will due to this also change.</p> + <p> + The previous strategy can still be enabled by passing the + <seealso marker="erl#+sws">+sws legacy</seealso> command + line flag to <c>erl</c>.</p> + <p> + Own Id: OTP-10661 Aux Id: OTP-10033 </p> + </item> + <item> + <p> + The <seealso marker="#+stbt">+stbt</seealso> command line + argument of <c>erl</c> was added. This argument can be + used for trying to set scheduler bind type. Upon failure + unbound schedulers will be used.</p> + <p> + Own Id: OTP-10668</p> + </item> + <item> + <p> + Support ANSI in console</p> + <p> + Unix platforms will no longer filter control sequences to + the ttsl driver thus enabling ANSI and colors in console. + (Thanks to Pedram Nimreezi)</p> + <p> + Own Id: OTP-10678</p> + </item> + <item> + <p>Add file:allocate/3 operation</p> + <p>This operation allows pre-allocation of space for + files. It succeeds only on systems that support such + operation. (Thanks to Filipe David Manana)</p> + <p> + Own Id: OTP-10680</p> + </item> + <item> + <p>Treat <c>-Wreturn-type</c> warnings as error when + using GCC (Thanks to Tuncer Ayaz)</p> + <p> + Own Id: OTP-10683</p> + </item> + <item> + <p> + Implement ./otp_build configure --enable-silent-rules</p> + <p> + With silent rules, the output of make is less verbose and + compilation warnings are easier to spot. Silent rules are + disabled by default and can be disabled or enabled at + will by make V=0 and make V=1. (Thanks to Anthony Ramine)</p> + <p> + Own Id: OTP-10726</p> + </item> + <item> + <p> + Use share flags for all file operations on Windows. + Thanks to Filipe David Borba Manana.</p> + <p> + Own Id: OTP-10727</p> + </item> + <item> + <p> + Make/fakefop adjustments. Thanks to Tuncer Ayaz and + Sebastian Rasmussen.</p> + <p> + Own Id: OTP-10733</p> + </item> + <item> + <p> + The runtime system will now by default use 10 async + threads if thread support has been enabled when building + the runtime system.</p> + <p> + This will prevent long blocking file-operations from + blocking scheduler threads for long periods of time, + which can be harmful. Apart from file-operations, it also + effects other operations scheduled on the async thread + pool by user implemented drivers.</p> + <p> + The amount of async threads can be controlled by using + the <seealso + marker="erl#async_thread_pool_size"><c>+A</c></seealso> + command line argument of <c>erl(1)</c>. When running some + offline tools you <em>might</em> want to disable async + threads, but you are advised <em>not</em> to in the + general case. Instead, you might want to increase the + amount of async threads used.</p> + <p> + This change imply changes of the characteristics the + system compared to the previous default. The + responsiveness of the system as a whole will be improved. + Operations scheduled on the async thread pool will get an + increased latency. The throughput of these operations may + increase, or decrease depending on the type of the + operations and how they get scheduled. In the case of + file operations, the throughput very much depends on how + the Erlang application access files. Multiple concurrent + accesses to different files have the potential of an + increased throughput.</p> + <p> + Own Id: OTP-10736</p> + </item> + <item> + <p> + The default reader group limit has been increased to 64 + from 8. This limit can be set using the <c>+rg</c> + command line argument of <c>erl(1)</c>.</p> + <p> + This change of default value will reduce lock contention + on ETS tables using the <c>read_concurrency</c> option at + the expense of memory consumption when the amount of + schedulers and logical processors are beween 8 and 64. + For more information, see documentation of the <c>+rg</c> + command line argument of <c>erl(1)</c>.</p> + <p> + Own Id: OTP-10737</p> + </item> + <item> + <p> + New BIF float_to_list/2 which solves a problem of + float_to_list/1 that doesn't allow specifying the number + of digits after the decimal point when formatting floats + (Thanks to Serge Aleynikov).</p> + <p> + Own Id: OTP-10752</p> + </item> + <item> + <p> + Limited support for unicode atoms in the external format + and in the internal representation of the vm. This is a + preparative feature in order to support communication + with future releases of Erlang/OTP that may create + unicode atoms.</p> + <p> + Own Id: OTP-10753</p> + </item> + <item> + <p> + Increased potential concurrency in ETS for + <c>write_concurrency</c> option. The number of internal + table locks has increased from 16 to 64. This makes it + four times less likely that two concurrent processes + writing to the same table would collide and thereby + serialized. The cost is an increased constant memory + footprint for tables using write_concurrency. The memory + consumption per inserted record is not affected. The + increased footprint can be particularly large if + <c>write_concurrency</c> is combined with + <c>read_concurrency</c>.</p> + <p> + Own Id: OTP-10787</p> + </item> + </list> + </section> + +</section> + <section><title>Erts 5.9.3.1</title> <section><title>Known Bugs and Problems</title> |