diff options
Diffstat (limited to 'erts/doc/src/notes.xml')
-rw-r--r-- | erts/doc/src/notes.xml | 1652 |
1 files changed, 1647 insertions, 5 deletions
diff --git a/erts/doc/src/notes.xml b/erts/doc/src/notes.xml index 1b91e5bc58..7021da7fb3 100644 --- a/erts/doc/src/notes.xml +++ b/erts/doc/src/notes.xml @@ -4,7 +4,7 @@ <chapter> <header> <copyright> - <year>2004</year><year>2015</year> + <year>2004</year><year>2017</year> <holder>Ericsson AB. All Rights Reserved.</holder> </copyright> <legalnotice> @@ -31,6 +31,1648 @@ </header> <p>This document describes the changes made to the ERTS application.</p> + +<section><title>Erts 8.3.3</title> + + <section><title>Fixed Bugs and Malfunctions</title> + <list> + <item> + <p> + Fixed memory segment cache used for multiblock carriers. + Huge (> 2GB) memory segments could cause a VM crash. + Creation of such huge memory segments used for multiblock + carriers is however very uncommon.</p> + <p> + Own Id: OTP-14360 Aux Id: ERL-401, PR-1417 </p> + </item> + <item> + <p> + Fix release note for OTP-14290 in ERTS version 8.3.1. It + was erroneously placed under "Known Bugs and Problems".</p> + <p> + Own Id: OTP-14363 Aux Id: OTP-14290 </p> + </item> + </list> + </section> + +</section> + +<section><title>Erts 8.3.2</title> + + <section><title>Fixed Bugs and Malfunctions</title> + <list> + <item> + <p> + The <c>+Bi</c> command line argument of <c>erl</c> + erroneously caused <c>SIGTERM</c> to be ignored by the VM + as well as of all its child processes. This bug was + introduced in erts version 8.3.</p> + <p> + Own Id: OTP-14358 Aux Id: OTP-14085 </p> + </item> + </list> + </section> + +</section> + +<section><title>Erts 8.3.1</title> + + <section><title>Fixed Bugs and Malfunctions</title> + <list> + <item> + <p> + Invoking <c>init:stop/0</c> via the SIGTERM signal, in a + non-SMP BEAM, could cause BEAM to terminate with fatal + error. This has now been fixed and the BEAM will + terminate normally when SIGTERM is received.</p> + <p> + Own Id: OTP-14290</p> + </item> + <item> + <p> + Trying to open a directory with file:read_file/1 on Unix + leaked a file descriptor. This bug has now been fixed.</p> + <p> + Own Id: OTP-14308 Aux Id: ERL-383 </p> + </item> + </list> + </section> + +</section> + +<section><title>Erts 8.3</title> + + <section><title>Fixed Bugs and Malfunctions</title> + <list> + <item> + <p>Fixed a number of bugs that caused faulty stack-traces + to be generated. The faulty stack traces were generated + either when applying the following functions or tracing + the following functions:</p> <list> + <item><c>erlang:error/1</c></item> + <item><c>erlang:error/2</c></item> + <item><c>erlang:exit/1</c></item> + <item><c>erlang:throw/1</c></item> </list> + <p> + Own Id: OTP-14055</p> + </item> + <item> + <p> + Corrected documentation about memory footprint for maps.</p> + <p> + Own Id: OTP-14118</p> + </item> + <item> + <p> + Fix <c>process_info(Pid, current_stacktrace)</c> to use + stack depth limit set by + <c>system_flag(backtrace_depth)</c>. The old behavior was + a hard coded depth limit of 8.</p> + <p> + Own Id: OTP-14119 Aux Id: PR-1263 </p> + </item> + <item> + <p> + A process calling <seealso + marker="erlang#system_flag_multi_scheduling"><c>erlang:system_flag(multi_scheduling, + block)</c></seealso> could end up hanging forever in the + call.</p> + <p> + Own Id: OTP-14121</p> + </item> + <item> + <p>Dirty scheduler bug fixes:</p> <list> <item><p>Fixed + call time tracing of process being scheduled on dirty + scheduler.</p></item> <item><p>GC info from dirty + schedulers.</p></item> <item><p>Multi scheduling block + with dirty schedulers could crash the runtime + system.</p></item> <item><p>Process structures could be + removed prematurely.</p></item> <item><p>GC on dirty + scheduler could crash the runtime system.</p></item> + <item><p>Termination of a process executing on a dirty + scheduler could cause a runtime system crash.</p></item> + </list> + <p> + Own Id: OTP-14122</p> + </item> + <item> + <p> + Fixed crash that occurred when writing timer data to a + crash dump.</p> + <p> + Own Id: OTP-14133</p> + </item> + <item> + <p> + A literal area could be removed while still referred from + processes.</p> + <p> + Own Id: OTP-14134</p> + </item> + <item> + <p> + Fixed a bug in the garbage collector that could crash the + runtime system.</p> + <p> + Own Id: OTP-14135</p> + </item> + <item> + <p> + Fixed a bug in call-time trace for NIFs which caused + tracing to erroneously be started multiple times for one + call.</p> + <p> + Own Id: OTP-14136</p> + </item> + <item> + <p> + Remove a debug printout and an unnecessary garbage + collection when handling exceptions in hipe compiled + code.</p> + <p> + Own Id: OTP-14153</p> + </item> + <item> + <p> + Fix bug in tracing of garbage collection that could cause + VM crash. Bug exists since OTP 19.0.</p> + <p> + Own Id: OTP-14154</p> + </item> + <item> + <p> + Fix bug in <c>binary_to_term</c> for binaries created by + <c>term_to_binary </c> with option <c>compressed</c>. The + bug can cause <c>badarg</c> exception for a valid binary + when Erlang VM is linked against a <c>zlib</c> library of + version 1.2.9 or newer. Bug exists since OTP 17.0.</p> + <p> + Own Id: OTP-14159 Aux Id: ERL-340 </p> + </item> + <item> + <p> + Fix suspension of schedulers when generating a crashdump.</p> + <p> + Own Id: OTP-14164</p> + </item> + <item> + <p>NIF resources was not handled in a thread-safe manner + in the runtime system without SMP support.</p> + <p>As a consequence of this fix, the following driver + functions are now thread-safe also in the runtime system + without SMP support:</p> <list> + <item><p><c>driver_free_binary()</c></p></item> + <item><p><c>driver_realloc_binary()</c></p></item> + <item><p><c>driver_binary_get_refc()</c></p></item> + <item><p><c>driver_binary_inc_refc()</c></p></item> + <item><p><c>driver_binary_dec_refc()</c></p></item> + </list> + <p> + Own Id: OTP-14202</p> + </item> + <item> + <p> + Fix <c>erlang:round/1</c> for large floating point + numbers with an odd absolute value between <c>(1 bsl + 52)</c> and <c>(1 bsl 53)</c>. The result was falsely + calculated as the next higher even number even though all + integer values up to <c>(1 bsl 53)</c> can be represented + as floats with full precision.</p> + <p> + Own Id: OTP-14227</p> + </item> + <item> + <p> + Add size of literals to module code size in crash dump + and <c>(l)oaded</c> command in break menu like it used to + be before OTP-19.0.</p> + <p> + Own Id: OTP-14228</p> + </item> + <item> + <p> + Fix potential bug in <c>enif_send</c> when called without + a process context and with argument <c>msg_env</c> as + <c>NULL</c>.</p> + <p> + Own Id: OTP-14229</p> + </item> + <item> + <p> + Fix bug where passing an appendable binary to + <c>erlang:port_control()</c> could crash the emulator.</p> + <p> + Own Id: OTP-14231</p> + </item> + <item> + <p> + Receive expressions with timeout in the Erlang shell + could cause a VM crash.</p> + <p> + Own Id: OTP-14241 Aux Id: ERL-365 </p> + </item> + </list> + </section> + + + <section><title>Improvements and New Features</title> + <list> + <item> + <p> + A received SIGTERM signal to beam will generate a + <c>'stop'</c> message to the <c>init</c> process and + terminate the Erlang VM nicely. This is equivalent to + calling <c>init:stop/0</c>.</p> + <p> + Own Id: OTP-14085</p> + </item> + <item> + <p> + Workaround for buggy Android implementation of + <c>PTHREAD_STACK_MIN</c> causing build of runtime system + to crash on undeclared <c>PAGE_SIZE</c>.</p> + <p> + Own Id: OTP-14165 Aux Id: ERL-319 </p> + </item> + <item> + <p> + Add configure option --without-thread-names that removes + the naming of individual emulator threads.</p> + <p> + Own Id: OTP-14234</p> + </item> + <item> + <p> + Add warning in documentation of <c>zlib:deflateInit/6</c> + about option <c>WindowsBits</c> values 8 and -8.</p> + <p> + Own Id: OTP-14254 Aux Id: ERL-362 </p> + </item> + </list> + </section> + +</section> + +<section><title>Erts 8.2.2</title> + + <section><title>Fixed Bugs and Malfunctions</title> + <list> + <item> + <p> + Fix bug in <c>binary_to_term</c> for binaries created by + <c>term_to_binary </c> with option <c>compressed</c>. The + bug can cause <c>badarg</c> exception for a valid binary + when Erlang VM is linked against a <c>zlib</c> library of + version 1.2.9 or newer. Bug exists since OTP 17.0.</p> + <p> + Own Id: OTP-14159 Aux Id: ERL-340 </p> + </item> + <item> + <p> + The driver efile_drv when opening a file now use fstat() + on the open file instead of stat() before opening, if + fstat() exists. This avoids a race when the file happens + to change between stat() and open().</p> + <p> + Own Id: OTP-14184 Aux Id: seq-13266 </p> + </item> + </list> + </section> + +</section> + +<section><title>Erts 8.2.1</title> + <section><title>Fixed Bugs and Malfunctions</title> + <list> + <item> + <p> + Fix a quite rare bug causing VM crash during code loading + and the use of export funs (fun M:F/A) of not yet loaded + modules. Requires a very specfic timing of concurrent + scheduler threads. Has been seen on ARM but can probably + also occure on other architectures. Bug has existed since + OTP R16.</p> + <p> + Own Id: OTP-14144 Aux Id: seq13242 </p> + </item> + </list> + </section> + +</section> + +<section><title>Erts 8.2</title> + + <section><title>Fixed Bugs and Malfunctions</title> + <list> + <item> + <p> + Fixed <c>configure</c> failures on MacOSX. Most important + <c>clock_gettime()</c> was detected when building for + MacOSX - El Capitan using XCode 8 despite it is not + available until MacOSX - Sierra.</p> + <p> + Own Id: OTP-13904 Aux Id: ERL-256 </p> + </item> + <item> + <p> + <c>code:add_pathsa/1</c> and command line option + <c>-pa</c> both revert the given list of directories when + adding it at the beginning of the code path. This is now + documented.</p> + <p> + Own Id: OTP-13920 Aux Id: ERL-267 </p> + </item> + <item> + <p> + Fix a compilation error of erts in OpenBSD related to the + usage of the __errno variable.</p> + <p> + Own Id: OTP-13927</p> + </item> + <item> + <p> + Fixed so that when enabling tracing on a process that had + an invalid tracer associated with it, the new tracer + overwrites the old tracer. Before this fix, calling + erlang:trace/3 would behave as if the tracer was still + alive and not apply the new trace.</p> + <p> + This fault was introduced in ERTS 8.0.</p> + <p> + Own Id: OTP-13928</p> + </item> + <item> + <p> + Fix parsing of <c>-profile_boot 'true' | 'false'</c></p> + <p> + Own Id: OTP-13955 Aux Id: ERL-280 </p> + </item> + <item> + <p> + A slight improvement of <c>erlang:get_stacktrace/0</c> + for exceptions raised in hipe compiled code. Beam + compiled functions in such stack trace was earlier + replaced by some unrelated function. They are now instead + omitted. This is an attempt to reduce the confusion in + the absence of a complete and correct stack trace for + mixed beam and hipe functions.</p> + <p> + Own Id: OTP-13992</p> + </item> + <item> + <p> Correct type declaration of match specification head. + </p> + <p> + Own Id: OTP-13996</p> + </item> + <item> + <p> + HiPE code loading failed for x86_64 if gcc was configured + with <c>--enable-default-pie</c>. Fixed by disabling PIE, + if needed for HiPE, when building the VM.</p> + <p> + Own Id: OTP-14031 Aux Id: ERL-294, PR-1239 </p> + </item> + <item> + <p> + Faulty arguments could be presented on exception from a + NIF that had rescheduled itself using + <c>enif_schedule_nif()</c>.</p> + <p> + Own Id: OTP-14048</p> + </item> + <item> + <p> + The runtime system could crash if a garbage collection on + a process was performed immediately after a NIF had been + rescheduled using <c>enif_schedule_nif()</c>.</p> + <p> + Own Id: OTP-14049</p> + </item> + <item> + <p> + A reference to purged code could be left undetected by + the purge operation if a process just had rescheduled a + NIF call using <c>enif_schedule_nif()</c> when the + process was checked. This could cause a runtime system + crash.</p> + <p> + Own Id: OTP-14050</p> + </item> + <item> + <p>Fixed a number of dirty scheduler related bugs:</p> + <list> <item><p>Process priority was not handled correct + when scheduling on a dirty scheduler.</p></item> + <item><p>The runtime system could crash when an exit + signal with a compound exit reason was sent to a process + executing on a dirty scheduler.</p></item> <item><p>The + runtime system crashed when call tracing a process + executing on a dirty scheduler.</p></item> <item><p>A + code purge operation could end up hanging forever when a + process executed on a dirty scheduler</p></item> </list> + <p> + Own Id: OTP-14051</p> + </item> + </list> + </section> + + + <section><title>Improvements and New Features</title> + <list> + <item> + <p> + Fix minor soft purge race bug that could incorrectly + trigger code_server to load new code for the module if + the soft purge failed and no current version of the + module existed.</p> + <p> + Own Id: OTP-13925</p> + </item> + <item> + <p> + To ease troubleshooting, <c>erlang:load_nif/2</c> now + includes the return value from a failed call to + load/reload/upgrade in the text part of the error tuple. + The <c>crypto</c> NIF makes use of this feature by + returning the source line where/if the initialization + fails.</p> + <p> + Own Id: OTP-13951</p> + </item> + <item> + <p> + New environment variable <c>ERL_CRASH_DUMP_BYTES</c> can + be used to limit the size of crash dumps. If the limit is + reached, crash dump generation is aborted and the + generated file will be truncated.</p> + <p> + Own Id: OTP-14046</p> + </item> + </list> + </section> + +</section> + +<section><title>Erts 8.1.1</title> + + <section><title>Fixed Bugs and Malfunctions</title> + <list> + <item> + <p> + The emulator got a dynamic library dependency towards + libsctp, which on Linux was not intended since the + emulator there loads and resolves the needed sctp + functions in runtime. This has been fixed and a configure + switch --enable-sctp=lib has been added for those who + want such a library dependency.</p> + <p> + Own Id: OTP-13956 Aux Id: ERL-262, ERL-133 </p> + </item> + <item> + <p> + Fix SIGUSR1 crashdump generation</p> + <p> + Do not generate a core when a crashdump is asked for.</p> + <p> + Own Id: OTP-13997</p> + </item> + <item> + <p>The new functions in <c>code</c> that allows loading + of many modules at once had a performance problem. While + executing a helper function in the <c>erl_prim_loader</c> + process, garbage messages were produced. The garbages + messages were ignored and ultimately discarded, but there + would be a negative impact on performance and memory + usage. The number of garbage message depended on both the + number of modules to be loaded and the length of the code + path.</p> + <p>The functions affected of this problem were: + <c>atomic_load/1</c>, <c>ensure_modules_loaded/1</c>, and + <c>prepare_loading/1</c>.</p> + <p> + Own Id: OTP-14009</p> + </item> + </list> + </section> + +</section> + +<section><title>Erts 8.1</title> + + <section><title>Fixed Bugs and Malfunctions</title> + <list> + <item> + <p> + Fix bug for calls from hipe code to BIFs that disable GC + while yielding. Has been causing Dialyzer crashes on ARM + (and presumably all other non-intel platforms).</p> + <p> + Own Id: OTP-13724 Aux Id: PR-1116 </p> + </item> + <item> + <p> + Fix a bug where changing the current working directory of + the VM would not change the current working directory of + programs spawned using <c>erlang:open_port({spawn,""}, + ...)</c>.</p> + <p> + Own Id: OTP-13733 Aux Id: ERL-175 </p> + </item> + <item> + <p> + Fix a bug where disabling tracing from a process that had + return_to tracing enabled and was tracing on + <c>erlang:trace/3</c> would cause a segmentation fault.</p> + <p> + Own Id: OTP-13734</p> + </item> + <item> + <p> + Update all erts documentation to use simpler English, use + consistent terminology and be easier to navigate.</p> + <p> + Own Id: OTP-13740</p> + </item> + <item> + <p> + Add dirty schedulers to the microstate accounting + statistics.</p> + <p> + Own Id: OTP-13744</p> + </item> + <item> + <p> + Fixed dirty scheduler build support on 32-bit windows.</p> + <p> + Own Id: OTP-13759</p> + </item> + <item> + <p> + inet:getstat(Socket) on an SCTP socket returned 0 for + send stats. This bug has now been corrected. Reported by + systra as issue ERL-102 on bugs.erlang.org.</p> + <p> + Own Id: OTP-13773 Aux Id: ERL-102 </p> + </item> + <item> + <p> + AF_UNSPEC and unknown address families were misread by + UDP receive in prim_inet resulting in an exception. This + bug has now been corrected.</p> + <p> + Own Id: OTP-13775</p> + </item> + <item> + <p> + Sweep HiPE stack for literals during code purge.</p> + <p> + Own Id: OTP-13777 Aux Id: PR-1122 </p> + </item> + <item> + <p> + Fix bug in run_erl for OpenBSD that could cause it on + rare occations to exit without starting the program (erl) + at all.</p> + <p> + Own Id: OTP-13795</p> + </item> + <item> + <p> + Update build scripts to not make assumtions about where + env, cp and perl are located.</p> + <p> + Own Id: OTP-13800</p> + </item> + <item> + <p> + Fixed a bug where init:stop could deadlock if a process + with infinite shutdown timeout (e.g. a supervisor) + attempted to load code while terminating.</p> + <p> + Own Id: OTP-13802</p> + </item> + <item> + <p> + Fixed a segmentation fault on sparc CPUs when free'ing a + tracer module's state.</p> + <p> + Own Id: OTP-13803</p> + </item> + <item> + <p> + <c>fun</c>s was not properly handled during purge of a + module. This could cause a crash of the VM after a purge + of a module.</p> + <p> + Own Id: OTP-13809</p> + </item> + <item> + <p> + Fixed a memory leak when the process monitoring a port + crashed.</p> + <p> + Own Id: OTP-13818</p> + </item> + <item> + <p> + Fixed multiple dirty scheduler related tracing bugs.</p> + <p> + Own Id: OTP-13822</p> + </item> + <item> + <p> + Fix error handling in beam code runtime loader for a + number of cases when index and size fields got corrupted + (negative) values.</p> + <p> + Own Id: OTP-13848 Aux Id: ERL-216 </p> + </item> + <item> + <p> + Minor fix of dirty scheduler implementation.</p> + <p> + Own Id: OTP-13852</p> + </item> + <item> + <p> + Calls to <c>erl_drv_send_term()</c> or + <c>erl_drv_output_term()</c> from a non-scheduler thread + while the corresponding port was invalid caused the + emulator to enter an inconsistent state which eventually + caused an emulator crash.</p> + <p> + Own Id: OTP-13866</p> + </item> + <item> + <p> + Fix a rare race condition in <c>erlang:open_port({spawn, + ""}, ...)</c> that would result in the erl_child_setup + program aborting and cause the emulator to exit.</p> + <p> + Own Id: OTP-13868</p> + </item> + <item> + <p>Driver and NIF operations accessing processes or ports + could cause an emulator crash when used from + non-scheduler threads. Those operations are:</p> <list> + <item><c>erl_drv_send_term()</c></item> + <item><c>driver_send_term()</c></item> + <item><c>erl_drv_output_term()</c></item> + <item><c>driver_output_term()</c></item> + <item><c>enif_send()</c></item> + <item><c>enif_port_command()</c></item> </list> + <p> + Own Id: OTP-13869</p> + </item> + <item> + <p> + Fix start scripts generation dependency in Makefile</p> + <p> + Own Id: OTP-13871 Aux Id: ERL-241 </p> + </item> + <item> + <p> + The VM could crash if <c>erlang:get_stacktrace()</c> was + called after a rescheduled NIF had thrown an exception.</p> + <p> + Own Id: OTP-13877</p> + </item> + <item> + <p>Calling <c>code:delete/1</c> before a loading a module + with an on_load function, the old code would be + overwritten, causing a memory or a crash if NIFs were + involved. (Thanks to vans163 for reporting this bug.)</p> + <p> + Own Id: OTP-13893 Aux Id: ERL-240 </p> + </item> + </list> + </section> + + + <section><title>Improvements and New Features</title> + <list> + <item> + <p> + Improved accuracy of timeouts on MacOS X. This by setting + premature timeouts followed by a short actual timeout + during scheduler wait.</p> + <p> + Own Id: OTP-13698</p> + </item> + <item> + <p>Added the following symbolic time unit representations + to the <seealso + marker="erlang#type-time_unit"><c>erlang:time_unit()</c></seealso> + type:</p> <list> <item><c>second</c></item> + <item><c>millisecond</c></item> + <item><c>microsecond</c></item> + <item><c>nanosecond</c></item> </list> <p>The following + symbolic time unit representations are now + <em>deprecated</em>, but still part of the + <c>erlang:time_unit()</c> type:</p> <list> + <item><c>seconds</c></item> + <item><c>milli_seconds</c></item> + <item><c>micro_seconds</c></item> + <item><c>nano_seconds</c></item> </list> + <p> + Own Id: OTP-13735</p> + </item> + <item> + <p> + Fix maps hashing entropy of maps with maps keys.</p> + <p> + Own Id: OTP-13763 Aux Id: ERL-199 </p> + </item> + <item> + <p> + Improved dirty scheduler support. A purge of a module can + now be performed without having to wait for completion of + all ongoing dirty NIF calls.</p> + <p> + Note that when enabling support for dirty schedulers, a + new purge strategy will as of ERTS version 8.1 be + enabled. This new strategy is not fully backwards + compatible with the strategy used by default. For more + information see the documentation of <seealso + marker="erts:erlang#check_process_code/3"><c>erlang:check_process_code/3</c></seealso>.</p> + <p> + Own Id: OTP-13808 Aux Id: OTP-13833 </p> + </item> + <item> + <p> + A new purge strategy has been introduced. The new + strategy will by default be disabled during the OTP 19 + release, but will be the only strategy available as of + the OTP 20 release.</p> + <p> + The new strategy is slightly incompatible with the + strategy being used by default in OTP 19. Using the + default strategy, processes holding <c>fun</c>s that + refer to the module being purged either fail a soft + purge, or will be killed during a hard purge. The new + strategy completely ignores <c>fun</c>s. If <c>fun</c>s + referring to the code being purged exist, and are used + after a purge, an exception will be raised upon usage. + That is, the behavior will be exactly the same as the + case when a <c>fun</c> is received by a process after the + purge.</p> + <p> + The new strategy can optionally be enabled when building + OTP during OTP 19, and will automatically be enabled if + the runtime system is built with support for dirty + schedulers.</p> + <p> + For more information see the documentation of <seealso + marker="erts:erlang#check_process_code/3"><c>erlang:check_process_code/3</c></seealso>.</p> + <p> + Own Id: OTP-13833</p> + </item> + <item> + <p> + Fixed unnecessary overestimation of heap size need during + garbage collection.</p> + <p> + Own Id: OTP-13851</p> + </item> + </list> + </section> + +</section> + +<section><title>Erts 8.0.5</title> + + <section><title>Fixed Bugs and Malfunctions</title> + <list> + <item> + <p> + Fixed a VM crash that occured in a garbage collection of + a process when it had received binaries. This bug was + introduced in ERTS version 8.0 (OTP 19.0).</p> + <p> + Own Id: OTP-13890</p> + </item> + </list> + </section> + +</section> + +<section><title>Erts 8.0.4</title> + + <section><title>Fixed Bugs and Malfunctions</title> + <list> + <item> + <p> + Fixed a VM crash that occured in garbage collection of a + process when it had received maps over the distribution. + This bug was introduced in ERTS version 8.0 (OTP 19.0).</p> + <p> + Own Id: OTP-13889</p> + </item> + </list> + </section> + +</section> + +<section><title>Erts 8.0.3</title> + + <section><title>Fixed Bugs and Malfunctions</title> + <list> + <item> + <p> + Fixed a race that could cause a lost wakeup of a process + that timed out in a <c>receive ... after</c>. This bug + was introduced in ERTS version 7.0.</p> + <p> + Own Id: OTP-13798 Aux Id: OTP-11997 </p> + </item> + <item> + <p> + Fixed segfault after writing an erl crash dump.</p> + <p> + Own Id: OTP-13799</p> + </item> + </list> + </section> + +</section> + +<section><title>Erts 8.0.2</title> + + <section><title>Fixed Bugs and Malfunctions</title> + <list> + <item> + <p> + Fix scheduler deadlock bug in <c>ets:update_counter/4</c> + when key is not found and inserting the default object + causes the table to grow.</p> + <p> + Own Id: OTP-13731 Aux Id: ERL-188 </p> + </item> + <item> + <p> + Fix VM abort "Overrun stack and heap" in garbage + collection triggered by a <c>bsl</c> operation and some + very specific heap conditions.</p> + <p> + Own Id: OTP-13732 Aux Id: seq13142 </p> + </item> + </list> + </section> + +</section> + +<section><title>Erts 8.0.1</title> + + <section><title>Fixed Bugs and Malfunctions</title> + <list> + <item> + <p> + A memory allocation bug in <c>group_leader/2</c> could + cause an emulator crash when garbage collecting a process + that had been assigned a remote group leader. This bug + was introduced in ERTS version 8.0.</p> + <p> + Own Id: OTP-13716</p> + </item> + </list> + </section> + +</section> + +<section><title>Erts 8.0</title> + + <section><title>Fixed Bugs and Malfunctions</title> + <list> + <item> + <p>The handling of <c>on_load</c> functions has been + improved. The major improvement is that if a code upgrade + fails because the <c>on_load</c> function fails, the + previous version of the module will now be retained.</p> + <p> + Own Id: OTP-12593</p> + </item> + <item> + <p><c>is_builtin(erlang, apply, 3)</c> will now return + <c>true</c>.</p> + <p> + Own Id: OTP-13034</p> + </item> + <item> + <p> + Fix <c>enif_get_list_length</c> to return false if list + is improper or have length larger than <c>UINT_MAX</c> + (did return true and an incorrect length value).</p> + <p> + Own Id: OTP-13288 Aux Id: PR913 </p> + </item> + <item> + <p> + Cleanup hipe signal handling code for x86 and make it + more portable.</p> + <p> + Own Id: OTP-13341 Aux Id: PR951 </p> + </item> + <item> + <p> + Make file:datasync use fsync instead of fdatasync on Mac + OSX.</p> + <p> + Own Id: OTP-13411</p> + </item> + <item> + <p> + Make sure to create a crash dump when running out of + memory. This was accidentally removed in the erts-7.3 + release.</p> + <p> + Own Id: OTP-13419</p> + </item> + <item> + <p> + A bug has been fixed where if erlang was started +B on a + unix platform it would be killed by a SIGUSR2 signal when + creating a crash dump.</p> + <p> + Own Id: OTP-13425</p> + </item> + <item> + <p> + Fix race between <c>process_flag(trap_exit,true)</c> and + a received exit signal.</p> + <p> + A process could terminate due to exit signal even though + <c>process_flag(trap_exit,true)</c> had returned. A very + specific timing between call to <c>process_flag/2</c> and + exit signal from another scheduler was required for this + to happen.</p> + <p> + Own Id: OTP-13452</p> + </item> + <item> + <p>Don't search for non-existing Map keys twice</p> + <p>For <c>maps:get/2,3</c> and <c>maps:find/2</c>, + searching for an immediate key, e.g. an atom, in a small + map, the search was performed twice if the key did not + exist.</p> + <p> + Own Id: OTP-13459</p> + </item> + <item> + <p> + When an abnormally large distribution message is about to + be sent, the VM has been changed to create a crash dump + instead of a core dump.</p> + <p> + Own Id: OTP-13474</p> + </item> + <item> + <p> + Fix <c>erlang:process_info/2</c> type specification</p> + <p> + Own Id: OTP-13485 Aux Id: ERL-123 </p> + </item> + <item> + <p> + Fix bug in <c>open_port/2</c> with option <c>{args, + List}</c>. A vm crash could be caused by an improper + <c>List</c>.</p> + <p> + Own Id: OTP-13489 Aux Id: ERL-127 </p> + </item> + <item> + <p> + Fixed a race-condition bug where the emulator could crash + when <c>erlang:system_profile/1,2</c> was enabled and a + process had to be re-scheduled during termination.</p> + <p> + Own Id: OTP-13494 Aux Id: ERL-126 </p> + </item> + <item> + <p> + Fixed bugs where the reduction counter was not handled + correct.</p> + <p> + Own Id: OTP-13512</p> + </item> + <item> + <p> + Fixed typo in description of the <c>EPMD_DUMP_REQ</c> + response.</p> + <p> + Own Id: OTP-13517</p> + </item> + <item> + <p> + Fixed a bug where a process flagged as sensitive would + sometimes record its save_calls when it shouldn't.</p> + <p> + Own Id: OTP-13540</p> + </item> + <item> + <p> + Update configure scripts to not use hard-coded path for + /bin/pwd and /bin/rm.</p> + <p> + Own Id: OTP-13562</p> + </item> + <item> + <p> + When passing a larger binary than the outputv callback of + a linked-in driver can handle in one io vector slot, the + binary is now split into multiple slots in the io vector. + This change only effects system where the max size of an + io vector slot is smaller then the word size of the + system (e.g. Windows).</p> + <p> + This change means that it is now possible on Windows to + send binaries that are larger than 4GB to port_command, + which is what is used for file:write, gen_tcp:send etc.</p> + <p> + Own Id: OTP-13628</p> + </item> + <item> + <p> + Workaround of Maps output in crashdumps. Currently the + atom 'undefined' is generated instead of Map data if a + Map type is encountered during crash.</p> + <p> + Own Id: OTP-13657</p> + </item> + </list> + </section> + + + <section><title>Improvements and New Features</title> + <list> + <item> + <p> + The tracing support has been extended to allow a <seealso + marker="erl_tracer">tracer module</seealso> to be the + trace event handler instead of a process or port. The + <seealso marker="erl_tracer">tracer module</seealso> + makes it possible for trace tools to filter or manipulate + trace event data without the trace event first having to + be copied from the traced process or port.</p> + <p> + With the introduction of this feature, + <c>erlang:trace(all|existing, _, _)</c> now also returns + the tracer process as part of the number of processes on + which tracing is enabled. The is incompatible with the + previous releases.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-10267</p> + </item> + <item> + <p> + Introduce LTTng tracing of Erlang Runtime System</p> + <p> + For LTTng to be enabled OTP needs to be built with + configure option <c>--with-dynamic-trace=lttng</c>.</p> + <p> + This feature introduces tracepoints for schedulers, + drivers, memory carriers, memory and async thread pool.</p> + <p> For a list of all tracepoints, see <seealso + marker="runtime_tools:LTTng">Runtime Tools User's + Guide</seealso> .</p> + <p> + Own Id: OTP-10282</p> + </item> + <item> + <p> + Make it possible to monitor/demonitor ports using the + <seealso + marker="erlang#monitor/2">erlang:monitor/2</seealso> API. + The process and port information functions have also been + updated to include information about monitors from + processes to ports.</p> + <p> + Own Id: OTP-11384</p> + </item> + <item> + <p> + Add microstate accounting</p> + <p> + Microstate accounting is a way to track which state the + different threads within ERTS are in. The main usage area + is to pin point performance bottlenecks by checking which + states the threads are in and then from there figuring + out why and where to optimize.</p> + <p> + Since checking whether microstate accounting is on or off + is relatively expensive only a few of the states are + enabled by default and more states can be enabled through + configure.</p> + <p> + There is a convenience module called msacc that has been + added to runtime_tools that can assist in gathering and + interpreting the data from Microstate accounting.</p> + <p> + For more information see <seealso + marker="erts:erlang#statistics_microstate_accounting">erlang:statistics(microstate_accounting, + _)</seealso> and the <seealso + marker="runtime_tools:msacc">msacc</seealso> module in + runtime_tools.</p> + <p> + Own Id: OTP-12345</p> + </item> + <item> + <p> + The port of Erlang/OTP to the real-time operating system + OSE has been removed.</p> + <p> + Own Id: OTP-12573</p> + </item> + <item> + <p> + Sharing preserved copy for messages and exit signals</p> + <p> + Enable sharing preserved copy with configure option + <c>--enable-sharing-preserving</c>. This will preserve + sharing, within the process, when communication with + other processes in the Erlang node. There is a trade-off, + the copy is more costly but this cost can be reclaimed if + there is a lot of sharing in the message. In addition + literals will not be copied in a send except during a + purge phase of the module where the literals are located. + This feature is considered experimental in 19.0.</p> + <p> + Own Id: OTP-12590 Aux Id: OTP-10251 </p> + </item> + <item> + <p> + Halfword BEAM has been removed.</p> + <p> + Own Id: OTP-12883</p> + </item> + <item> + <p> + Added <seealso + marker="kernel:os#perf_counter/1">os:perf_counter/1</seealso>.</p> + <p> + The perf_counter is a very very cheap and high resolution + timer that can be used to timestamp system events. It + does not have monoticity guarantees, but should on most + OS's expose a monotonous time.</p> + <p> + Own Id: OTP-12908</p> + </item> + <item> + <p> + Support for a fragmented young heap generation. That is, + the young heap generation can consist of multiple non + continuous memory areas. The main reason for this change + is to avoid extra copying of messages that could not be + allocated directly on the receivers heap.</p> + <p> + Own Id: OTP-13047</p> + </item> + <item> + <p> + Erlang linked-in driver can now force the call to + open_port to block until a call to erl_drv_init_ack is + made inside the driver. This is useful when you want to + do some asynchronous initialization, for example getting + configuration from a pipe, and you want the initial + open_port call to fail if the configuration is incomplete + or wrong. See the erl_driver documentation for more + details on the API.</p> + <p> + Own Id: OTP-13086</p> + </item> + <item> + <p> + Erlang linked-in drivers can now set their own pids as + seen in <c>erlang:port_info/1</c> by using the + <c>erl_drv_set_pid</c> function. For more details see the + erl_driver documentation.</p> + <p> + Own Id: OTP-13087</p> + </item> + <item> + <p> + The functionality behind <c>erlang:open_port/2</c> when + called with spawn or spawn_executable has been redone so + that the forking of the new program is done in a separate + process called erl_child_setup. This allows for a much + more robust implementation that uses less memory and does + not block the entire emulator if the program to be + started is on an un-accessible NFS. Benchmarks have shown + this approach to be about 3-5 times as fast as the old + approach where the fork/vfork was done by erts. This is a + pure stability and performance fix, however some error + messages may have changed, which is why it is marked as a + backwards incompatible change.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-13088</p> + </item> + <item> + <p>Improved yielding strategy in the implementation of + the following native functions:</p> <list> + <item><c>erlang:binary_to_list/1</c></item> + <item><c>erlang:binary_to_list/3</c></item> + <item><c>erlang:bitstring_to_list/1</c></item> + <item><c>erlang:list_to_binary/1</c></item> + <item><c>erlang:iolist_to_binary/1</c></item> + <item><c>erlang:list_to_bitstring/1</c></item> + <item><c>binary:list_to_bin/1</c></item> </list> <p>This + in order to improve performance of these functions.</p> + <p> + Own Id: OTP-13096</p> + </item> + <item> + <p> + All garbage collections of processes now bump reductions. + Also the amount of reductions bumped when garbage + collecting has been adjusted. It now better corresponds + to the amount of work performed. This in order to improve + the real time characteristics of the system.</p> + <p> + Own Id: OTP-13097</p> + </item> + <item> + <p>New functions that can load multiple modules at once + have been added to the '<c>code</c>' module. The + functions are <c>code:atomic_load/1</c>, + <c>code:prepare_loading/1</c>, + <c>code:finish_loading/1</c>, and + <c>code:ensure_modules_loaded/1</c>.</p> + <p> + Own Id: OTP-13111</p> + </item> + <item> + <p>The <c>-boot_var</c> option for <c>erl</c> now only + supports a single key and single value (as documented). + The option used to allow multiple key/value pairs, but + that behavior was undocumented.</p> + <p>The function <c>erl_prim_loader:start/3</c> has been + removed. Its documentation has also been removed.</p> + <p>The undocumented and unsupported function + <c>erl_prim_loader:get_files/2</c> has been removed.</p> + <p> + Own Id: OTP-13112</p> + </item> + <item> + <p> + Low level BIF <c>erlang:purge_module/1</c> is made more + robust against incorrect use. Lingering processes that + still refer the old code are now killed before the module + is purged to prevent fatal VM behavior.</p> + <p> + Own Id: OTP-13122</p> + </item> + <item> + <p>Improved dirty scheduler implementation. For more + information see the <seealso + marker="erl_nif#dirty_nifs">NIF + documentation</seealso>.</p> <note><list> <item><p>The + dirty scheduler support is still + <em>experimental</em>.</p></item> <item><p>The support + for determining whether dirty NIF support exist or not at + compile time using the C preprocessor macro + <c>ERL_NIF_DIRTY_SCHEDULER_SUPPORT</c> has been + removed.</p></item> <item><p>The + <c>enif_is_on_dirty_scheduler()</c> function has been + removed. Use <seealso + marker="erl_nif#enif_thread_type"><c>enif_thread_type()</c></seealso> + instead.</p></item> </list></note> + <p> + Own Id: OTP-13123</p> + </item> + <item> + <p> + Various optimizations done to process dictionary access.</p> + <p> + Own Id: OTP-13167</p> + </item> + <item> + <p> + Added max_heap_size process flag. max_heap_size allows + the user to limit the maximum heap used by a process. See + <seealso + marker="erlang#process_flag/2">erlang:process_flag</seealso> + for more details.</p> + <p> + Own Id: OTP-13174</p> + </item> + <item> + <p> + Allow dynamic drivers and NIF libraries to be built with + gcc option <c>-fvisibility=hidden</c> for faster loading + and more optimized code.</p> + <p> + Own Id: OTP-13227</p> + </item> + <item> + <p> + Add <c>erlang:process_info(Pid, + garbage_collection_info)</c> which returns extended + garbage_collection information. For more details see the + documentation.</p> + <p> + Own Id: OTP-13265</p> + </item> + <item> + <p> + The functions <c>erlang:list_to_integer/1</c> and + <c>string:to_integer/1</c> have been optimized for large + inputs.</p> + <p> + Own Id: OTP-13293</p> + </item> + <item> + <p> + Improved memory allocation strategy for hipe native code + on x86_64 (amd64) architectures by reserving enough low + virtual address space needed for the HiPE/AMD64 small + code model. The default virtual address area for hipe + code is set to 512Mb, but can be changed with emulator + flag <c>+MXscs</c>.</p> + <p> + Own Id: OTP-13359</p> + </item> + <item> + <p> + Introduction of configurable management of data referred + to by the message queue of a process. Each process can be + configured individually.</p> + <p> + It is now possible to configure the message queue of a + process, so that all data referred by it will be kept + outside of the heap, and by this prevent this data from + being part of garbage collections.</p> + <p> + For more information see the documentation of <seealso + marker="erlang#process_flag_message_queue_data"><c>process_flag(message_queue_data, + MQD)</c></seealso>.</p> + <p> + Own Id: OTP-13366 Aux Id: OTP-13047 </p> + </item> + <item> + <p> + Processes now yield when scanning large message queues + and not finding a matching message. This in order to + improve real time characteristics.</p> + <p> + Own Id: OTP-13401</p> + </item> + <item> + <p> + Optimized an erts internal function that is used to + traverse erlang terms. The internal function was mainly + used by term_to_binary and comparison of terms. + Benchmarks have shown up to a 10% performance increase in + those functions after the optimization.</p> + <p> + Own Id: OTP-13440</p> + </item> + <item> + <p>Add the following NIF API functions:</p> + <list> <item><seealso + marker="erl_nif#enif_cpu_time"><c>enif_cpu_time</c></seealso></item> + <item><seealso + marker="erl_nif#enif_now_time"><c>enif_now_time</c></seealso></item> + <item><seealso + marker="erl_nif#enif_make_unique_integer"><c>enif_make_unique_integer</c></seealso></item> + <item><seealso + marker="erl_nif#enif_is_process_alive"><c>enif_is_process_alive</c></seealso></item> + <item><seealso + marker="erl_nif#enif_is_port_alive"><c>enif_is_port_alive</c></seealso></item> + <item><seealso + marker="erl_nif#enif_term_to_binary"><c>enif_term_to_binary</c></seealso></item> + <item><seealso + marker="erl_nif#enif_binary_to_term"><c>enif_binary_to_term</c></seealso></item> + <item><seealso + marker="erl_nif#enif_port_command"><c>enif_port_command</c></seealso></item> + </list> + <p> + For details of what each function does, see the erl_nif + documentation.</p> + <p> + Own Id: OTP-13442</p> + </item> + <item> + <p> + Optimize <c>'++'</c> operator and <c>lists:append/2</c> + by using a single pass to build a new list while checking + for properness.</p> + <p> + Own Id: OTP-13487</p> + </item> + <item> + <p> + Handle terms (pids,ports and refs) from nodes with a + 'creation' value larger than 3. This is a preparation of + the distribution protocol to allow OTP 19 nodes to + correctly communicate with future nodes (20 or higher). + The 'creation' value differentiates different + incarnations of the same node (name).</p> + <p> + Own Id: OTP-13488</p> + </item> + <item> + <p> + Don't send unasked for systemd notifications in epmd</p> + <p> + Own Id: OTP-13493 Aux Id: PR-999 </p> + </item> + <item> + <p> + The enif_send API has been extended to allow NULL to be + used as the message environment. When used this way, a + message environment is implicitly created and the given + term is copied into that environment before sending. This + can be an optimization if many small messages are being + sent by the nif.</p> + <p> + Own Id: OTP-13495</p> + </item> + <item> + <p> + The tracing support has been extended to allow tracing on + ports. Ports can be traced on using the 'ports', 'send' + and 'receive' trace flags.</p> + <p> + The first argument of <seealso + marker="erts:erlang#trace/3">erlang:trace/3</seealso> has + been extended so that <c>'all'</c>, <c>'existing'</c> and + <c>'new'</c> now include both processes and ports. New + <c>Tracee</c> variants, <c>'all_processes'</c>, + <c>'all_ports'</c>, <c>'existing_processes'</c> etc have + been added to specify only processes or ports.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-13496</p> + </item> + <item> + <p> + When the <c>'procs'</c> trace flag is enabled, a + <c>'spawned'</c> trace event is now also generated by a + newly created process. The previous event <c>'spawn'</c> + remains, but as it is generated by the process that did + the spawn, it is not guaranteed that it is ordered with + other trace events from the newly spawned process. So + when tracking the lifetime of a process this new event + should be used as the creation event.</p> + <p> + This new trace event is marked as an incompatibility + because tools that expect certain trace events when + enabling 'procs' will have to updated.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-13497</p> + </item> + <item> + <p> + Add the <seealso + marker="erts:erlang#match_spec_test/3">erlang:match_spec_test/3</seealso> + function. The functions allows the testing of match + specifications for both tracing and ets tables. It can be + used to test that a match specification does the expected + filtering on specific data. It also returns more verbose + error reasons for incorrectly constructed match + specifications.</p> + <p> + Own Id: OTP-13501</p> + </item> + <item> + <p> + The erts internal tracing support has been changed to + have much less overhead and be more scalable.</p> + <p> + This rewrite does not break any backwards + incompatibilities, but it does change the ordering of + some trace messages when compared to previous releases. + It should be noted that this only applies to trace + messages sent to processes or ports, it does not apply to + the new tracer module. However in future releases they + may also be effected by this.</p> + <p> + Trace messages are only guaranteed to be ordered from one + traced process or port. In previous releases this was not + visible as a <c>'send'</c> trace message would always + arrive before the corresponding <c>'receive'</c> trace + message that is no longer always the case. This also + means that timestamped trace messages may seem to arrive + out of order as the timestamp is taken when the event is + triggered and not when it is put in the queue of the + tracer.</p> + <p> + Own Id: OTP-13503</p> + </item> + <item> + <p> + Add possibility to filter <c>send</c> and <c>receive</c> + trace with match specifications.</p> + <p> + Own Id: OTP-13507</p> + </item> + <item> + <p> + Add <c>maps:update_with/3,4</c> and <c>maps:take/2</c></p> + <p> + Own Id: OTP-13522 Aux Id: PR-1025 </p> + </item> + <item> + <p> + Introduce LTTng tracing via Erlang tracing.</p> + <p> + For LTTng to be enabled OTP needs to be built with + configure option <c>--with-dynamic-trace=lttng</c>.</p> + <p>The dynamic trace module <c>dyntrace</c> is now + capable to be used as a LTTng sink for Erlang tracing. + For a list of all tracepoints, see <seealso + marker="runtime_tools:LTTng">Runtime Tools User's + Guide</seealso> .</p> + <p>This feature also introduces an incompatible change in + trace tags. The trace tags <c>gc_start</c> and + <c>gc_end</c> has been split into <c>gc_minor_start</c>, + <c>gc_minor_end</c> and <c>gc_major_start</c>, + <c>gc_major_end</c>.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-13532</p> + </item> + <item> + <p> + Print heap pointers for garbing processes during + crashdump</p> + <p> + Own Id: OTP-13541 Aux Id: PR-1026 </p> + </item> + <item> + <p> + Changed and improved low level memory statistics returned + by <c>erlang:system_info/1</c>. The info for + <c>erts_mmap</c> has been moved from <c>mseg_alloc</c> to + its own section returned by <c>{allocator, + erts_mmap}</c>.</p> + <p> + Own Id: OTP-13560</p> + </item> + <item> + <p> + Add enif_snprintf to the NIF API</p> + <p> + The function <c>enif_snprintf</c> is similar to + <c>snprintf</c> call but can handle formatting of Erlang + terms via <c>%T</c> format specifier.</p> + <p> + Own Id: OTP-13580</p> + </item> + <item> + <p>The warning in the documentation for + <c>erlang:raise/3</c> has been removed. It is now + officially perfectly fine to use raise/3 in production + code.</p> + <p> + Own Id: OTP-13599</p> + </item> + <item> + <p> + Fix bugs caused by the VM sometimes truncating object + sizes or offsets to 32 bits on 64-bit hosts. These bugs + were mainly found when working with large unicode strings + and nifs environments.</p> + <p> + Own Id: OTP-13606</p> + </item> + <item> + <p> + Add <c>-start_epmd</c> command line option, this lets you + disable automatic starting of epmd when starting a + distributed node.</p> + <p> + Add <c>-epmd_module</c> command line option, this lets + you specify a module to register and look-up node names + in. The default module is <c>erl_epmd</c>.</p> + <p> + Own Id: OTP-13627</p> + </item> + <item> + <p> + <c>erlang:halt</c> now truncates strings longer than 200 + characters instead of failing with <c>badarg</c>.</p> + <p> + Own Id: OTP-13630</p> + </item> + <item> + <p> + Fix possible race in poller wake up on windows</p> + <p> + Own Id: OTP-13634</p> + </item> + </list> + </section> + +</section> + <section><title>Erts 7.3.1.3</title> <section><title>Fixed Bugs and Malfunctions</title> @@ -8694,9 +10336,9 @@ dynamically linking against <c>libssl.so</c> and <c>libcrypto.so</c>. The runtime library search path has also been extended. </item><item> The <c>configure</c> - scripts of <c>erl_interface</c> and <c>odbc</c> now + scripts of Erl_interface and ODBC now search for thread libraries and thread library quirks the - same way as <c>erts</c> do. </item><item> The + same way as ERTS do. </item><item> The <c>configure</c> script of the <c>odbc</c> application now also looks for odbc libraries in <c>lib64</c> and <c>lib/64</c> directories when building on a 64-bit @@ -9812,7 +11454,7 @@ </item> <item> <p> - A corrected bug in <c>ets</c> for <c>bag</c> and + A corrected bug in ETS for <c>bag</c> and <c>duplicate_bag</c>. A <c>delete/2</c> or <c>lookup_element/3</c> could miss objects in a fixed table if one or more objects with the same key had @@ -10254,7 +11896,7 @@ <list> <item> <p> - A corrected bug in <c>ets</c> for <c>bag</c> and + A corrected bug in ETS for <c>bag</c> and <c>duplicate_bag</c>. A <c>delete/2</c> or <c>lookup_element/3</c> could miss objects in a fixed table if one or more objects with the same key had |