aboutsummaryrefslogtreecommitdiffstats
path: root/lib/stdlib/doc
diff options
context:
space:
mode:
Diffstat (limited to 'lib/stdlib/doc')
-rw-r--r--lib/stdlib/doc/src/notes.xml241
1 files changed, 241 insertions, 0 deletions
diff --git a/lib/stdlib/doc/src/notes.xml b/lib/stdlib/doc/src/notes.xml
index ef12e525e3..605a9f224d 100644
--- a/lib/stdlib/doc/src/notes.xml
+++ b/lib/stdlib/doc/src/notes.xml
@@ -31,6 +31,247 @@
</header>
<p>This document describes the changes made to the STDLIB application.</p>
+<section><title>STDLIB 3.9</title>
+
+ <section><title>Fixed Bugs and Malfunctions</title>
+ <list>
+ <item>
+ <p> Fix a bug in <c>string:lexemes/2</c>. </p> <p> The
+ bug was found when optimizing the handling of deep lists
+ of Unicode characters in the <c>string</c> module. </p>
+ <p>
+ Own Id: OTP-15649</p>
+ </item>
+ <item>
+ <p>A bug has been fixed in the <c>maps</c> implementation
+ that could cause a crash or memory usage to grow until
+ the machine ran out of memory. This could happen when
+ inserting a new key-value pair with a key <c>K1</c>
+ containing a binary <c>B1</c> into a map <c>M</c> having
+ a key <c>K2</c> with a binary <c>B2</c> if the following
+ conditions were met:</p> <list> <item><c>B1 =/=
+ B2</c></item> <item><c>size(B1) >= 4294967296</c></item>
+ <item><c>size(B2) >= 4294967296</c></item>
+ <item><c>size(M) >= 32</c></item> <item><c>(size(B1) rem
+ 4294967296) == (size(B2) rem 4294967296)</c></item>
+ <item>the first <c>(size(B1) rem 4294967296)</c> bytes
+ are the same both in <c>B1</c> and <c>B2</c></item>
+ <item>substituting <c>B1</c> in <c>K1</c> with <c>B2</c>
+ would create a term with the same value as
+ <c>K2</c></item> </list> <p>The root cause of the problem
+ is that the <c>maps</c> implementation only hashed the
+ first <c>(X rem 4294967296)</c> bytes of binaries so that
+ different binaries could get the same hash value
+ independently of the hash seed.</p>
+ <p>
+ Own Id: OTP-15707</p>
+ </item>
+ <item>
+ <p> Since the introduction of the stack trace variable,
+ the Erlang Pretty Printer has left out the exception
+ class <c>throw</c> even when the stack trace variable
+ cannot be left out, which is not correct Erlang code. The
+ fix is to always include the exception class
+ <c>throw</c>. </p>
+ <p>
+ Own Id: OTP-15751</p>
+ </item>
+ <item>
+ <p><c>record_info/2</c> is a pseudo-function that
+ requires literal arguments known at compile time.
+ Therefore, the following usage is illegal: <c>fun
+ record/info/2</c>. The compiler would crash when during
+ compilation of that kind of code. Corrected to issue a
+ compilation error.</p>
+ <p>
+ Own Id: OTP-15760 Aux Id: ERL-907 </p>
+ </item>
+ </list>
+ </section>
+
+
+ <section><title>Improvements and New Features</title>
+ <list>
+ <item>
+ <p> A new <c>rand</c> module algorithm, <c>exro928ss</c>
+ (Xoroshiro928**), has been implemented. It has got a
+ really long period and good statistical quality for all
+ output bits, while still being only about 50% slower than
+ the default algorithm. </p><p> The same generator is also
+ used as a long period counter in a new <c>crypto</c>
+ plugin for the <c>rand</c> module, algorithm
+ <c>crypto_aes</c>. This plugin uses AES-256 to scramble
+ the counter which buries any detectable statistical
+ artifacts. Scrambling is done in chunks which are cached
+ to get good amortized speed (about half of the default
+ algorithm). </p>
+ <p>
+ Own Id: OTP-14461 Aux Id: PR-1857 </p>
+ </item>
+ <item>
+ <p>
+ Types related to server naming and starting have been
+ exported from <c>gen_statem</c>. These are:
+ <c>server_name/0</c>, <c>server_ref/0</c>,
+ <c>start_opt/0</c>, <c>start_ret/0</c> and
+ <c>enter_loop_opt/0</c>.</p>
+ <p>
+ Own Id: OTP-14724 Aux Id: PR-2056 </p>
+ </item>
+ <item>
+ <p>
+ The default algorithm for the <c>rand</c> module has been
+ changed to <c>exsss</c> (Xorshift116**) which is a
+ combination of the Xorshift116 (<c>exsp</c>) state update
+ and a new scrambler "StarStar" from the 2018 paper
+ "Scrambled Linear Pseudorandom Number Generators" by
+ David Blackman and Sebastiano Vigna. This combination
+ should not have the caveat of weak low bits that the
+ previous default algorithm(s) have had, with the cost of
+ about 10% lower speed. See GitHub pull request #1969.</p>
+ <p>
+ Own Id: OTP-14731 Aux Id: PR-1969 </p>
+ </item>
+ <item>
+ <p>
+ The generic state machine behaviour <c>gen_statem</c> has
+ gotten code cleanup and documentation improvements from
+ GitHub Pull Request #1855, even though the PR itself was
+ rejected.</p>
+ <p>
+ Own Id: OTP-14737 Aux Id: PR-1855 </p>
+ </item>
+ <item>
+ <p>
+ Update Unicode specification to version 11.0.</p>
+ <p>
+ Own Id: OTP-15111</p>
+ </item>
+ <item>
+ <p>
+ ETS option <c>write_concurrency</c> now also affects and
+ improves the scalability of <c>ordered_set</c> tables.
+ The implementation is based on a data structure called
+ contention adapting search tree, where the lock
+ granularity adapts to the actual amount of concurrency
+ exploited by the applications in runtime.</p>
+ <p>
+ Own Id: OTP-15128</p>
+ </item>
+ <item>
+ <p>
+ Optimized <c>maps:new/0</c> with trivial Erlang
+ implementation, making use of literal terms (the empty
+ map) not needing dynamic heap allocation.</p>
+ <p>
+ Own Id: OTP-15200 Aux Id: PR-1878 </p>
+ </item>
+ <item>
+ <p>The <c>gen_*</c> behaviours have been changed so that
+ if logging of the last N messages through
+ <c>sys:log/2,3</c> is active for the server, this log is
+ included in the terminate report.</p> <p>To accomplish
+ this the format of "System Events" as defined in the man
+ page for <c>sys</c> has been clarified and cleaned up, a
+ new function <c>sys:get_log/1</c> has been added, and
+ <c>sys:get_debug/3</c> has been deprecated. Due to these
+ changes, code that relies on the internal badly
+ documented format of "System Events", need to be
+ corrected.</p>
+ <p>
+ *** POTENTIAL INCOMPATIBILITY ***</p>
+ <p>
+ Own Id: OTP-15381</p>
+ </item>
+ <item>
+ <p>
+ The <c>gen_statem</c> behaviour engine loop has been
+ optimized for better performance in particular when the
+ callback module returns some actions, that is better
+ performance for more realistic applications than the Echo
+ Benchmark.</p>
+ <p>
+ Own Id: OTP-15452</p>
+ </item>
+ <item>
+ <p> Do not allow function specifications for functions
+ residing in other modules. </p>
+ <p>
+ *** POTENTIAL INCOMPATIBILITY ***</p>
+ <p>
+ Own Id: OTP-15563 Aux Id: ERL-845, OTP-15562 </p>
+ </item>
+ <item>
+ <p>
+ The <c>persistent_term</c> functions <c>put/2</c> and
+ <c>erase/1</c> are now yielding.</p>
+ <p>
+ Own Id: OTP-15615</p>
+ </item>
+ <item>
+ <p>Previously, all ETS tables used centralized counter
+ variables to keep track of the number of items stored and
+ the amount of memory consumed. These counters can cause
+ scalability problems (especially on big NUMA systems).
+ This change adds an implementation of a decentralized
+ counter and modifies the implementation of ETS so that
+ ETS tables of type <c>ordered_set</c> with
+ <c>write_concurrency</c> enabled use the decentralized
+ counter. Experiments indicate that this change
+ substantially improves the scalability of ETS
+ <c>ordered_set</c> tables with <c>write_concurrency</c>
+ enabled in scenarios with frequent <c>ets:insert/2</c>
+ and <c>ets:delete/2</c> calls.</p>
+ <p>
+ Own Id: OTP-15623 Aux Id: PR-2190 </p>
+ </item>
+ <item>
+ <p> Use <c>ssh</c> instead of <c>rsh</c> as the default
+ remote shell. </p>
+ <p>
+ Own Id: OTP-15633 Aux Id: PR-1787 </p>
+ </item>
+ <item>
+ <p>Added <c>beam_lib:strip/2</c> and friends, which
+ accept a list of chunks that should be preserved when
+ stripping.</p>
+ <p>
+ Own Id: OTP-15680 Aux Id: PR-2114 </p>
+ </item>
+ <item>
+ <p> Optimize printing of maps with <c>io_lib:write()</c>.
+ Also optimize pretty printing of strings (<c>~s</c> and
+ <c>~ts</c>) when limiting the output with the
+ <c>chars_limit</c> option. </p>
+ <p>
+ Own Id: OTP-15705</p>
+ </item>
+ <item>
+ <p> There are new compiler options <c>nowarn_removed</c>
+ and <c>{nowarn_removed,Items}</c> to suppress warnings
+ for functions and modules that have been removed from
+ OTP.</p>
+ <p>
+ Own Id: OTP-15749 Aux Id: ERL-904 </p>
+ </item>
+ <item>
+ <p> Let the Erlang Pretty Printer put atomic parts on the
+ same line. </p>
+ <p>
+ Own Id: OTP-15755</p>
+ </item>
+ <item>
+ <p> Add option <c>quote_singleton_atom_types</c> to the
+ Erlang Pretty Printer's functions. Setting the option to
+ <c>true</c> adds quotes to all singleton atom types. </p>
+ <p>
+ Own Id: OTP-15756</p>
+ </item>
+ </list>
+ </section>
+
+</section>
+
<section><title>STDLIB 3.8.2</title>
<section><title>Fixed Bugs and Malfunctions</title>