diff options
Diffstat (limited to 'erts/doc/src/notes.xml')
| -rw-r--r-- | erts/doc/src/notes.xml | 987 |
1 files changed, 985 insertions, 2 deletions
diff --git a/erts/doc/src/notes.xml b/erts/doc/src/notes.xml index 824ad6d94e..532ebc29e2 100644 --- a/erts/doc/src/notes.xml +++ b/erts/doc/src/notes.xml @@ -30,6 +30,989 @@ </header> <p>This document describes the changes made to the ERTS application.</p> +<section><title>Erts 5.8.1.1</title> + + <section><title>Fixed Bugs and Malfunctions</title> + <list> + <item> + <p> + Windows 2003 and Windows XP pre SP3 would sometimes not + start the Erlang R14B VM at all due to a bug in the cpu + topology detection. The bug affects Windows only, no + other platform is even remotely affected. The bug is now + corrected.</p> + <p> + Own Id: OTP-8876</p> + </item> + </list> + </section> + +</section> + +<section><title>Erts 5.8.1</title> + + <section><title>Fixed Bugs and Malfunctions</title> + <list> + <item> + <p> + Very small floating point numbers generated errors when + converting from list to float in some versions of the VM, + this is now corrected so that i.e. + list_to_float("1.0e-324"). returns 0.0 in all versions of + Erlang.</p> + <p> + Own Id: OTP-7178</p> + </item> + <item> + <p> + Windows Vista and Windows 7 file system virtualization, + which makes "old style" windows programs execute in a + file system sandbox, was previously unintentionally + turned on for the Erlang VM. This is now corrected so + that i.e. writes to C:\Program Files\. without + administrator privileges will fail.</p> + <p> + Own Id: OTP-7405</p> + </item> + <item> + <p> + Fix faulty 64-bit integer term output from drivers. Large + 64-bits integers did not generate correct bignums and + could even cause emulator crash. Only affects drivers + using ERL_DRV_INT64 or ERL_DRV_UINT64, introduced in + R13B03.</p> + <p> + Own Id: OTP-8716</p> + </item> + <item> + <p> + Fixed: inet:setopts(S, [{linger,{true,2}}]) returned + {error,einval} for SCTP sockets. The inet_drv had a bug + when checking the option size.</p> + <p> + Own Id: OTP-8726 Aux Id: seq11617 </p> + </item> + <item> + <p>Fix libm linking with --as-needed flag + <p> + When building with "--as-needed" linker flags on Linux + the build will fail. This has now been fixed.</p> + <p> + (Thanks to Christian Faulhammer)</p></p> + <p> + Own Id: OTP-8728</p> + </item> + <item> + <p> + gen_udp:connect/3 was broken for SCTP enabled builds. It + did not detect remote end errors as it should.</p> + <p> + Own Id: OTP-8729</p> + </item> + <item> + <p>Reduce the risk of integer wrapping in bin vheap size + counting.</p> <p>The vheap size series will now use the + golden ratio instead of doubling and fibonacci + sequences.</p> + <p> + Own Id: OTP-8730</p> + </item> + <item> + <p> + ETS ordered_set containing <c>[]</c> as key could cause + strange thing to happen, like an infinite hanging + <c>ets:select</c>.</p> + <p> + Own Id: OTP-8732</p> + </item> + <item> + <p>reference() has been substituted for ref() in the + documentation.</p> + <p> + Own Id: OTP-8733</p> + </item> + <item> + <p> + When a native compiled module called a not loaded + non-native compiled module that had an on_load function, + the export entries were trashed after code loading so on + the next call from the native compiled module to the + non-native compiled the emulator crashed. This bug has + now been fixed.</p> + <p> + Own Id: OTP-8736</p> + </item> + <item> + <p> + HiPE-enabled Erlang VMs running on BSD systems sometimes + generated messages like "Yikes! erts_alloc() returned + misaligned address 0x8016a512c". Fixed. (Thanks to Mikael + Pettersson.)</p> + <p> + Own Id: OTP-8769</p> + </item> + <item> + <p> + A race condition in <c>erts_poll()</c> could cause delay + of poll for I/O.</p> + <p> + Own Id: OTP-8773</p> + </item> + <item> + <p> + Removed some potential vulnerabilities from the Erlang + Port Mapper Daemon (epmd) and straightened up access + control. Also removed hazardous interfaces allowing + anyone on a machine to forcefully unregister other nodes. + This means that the ei_unregister/erl_unregister + interfaces in erl_interface is rendered not only error + prone and mystifying as before, but totally ineffective. + The old behaviour of unchecked node unregistering can be + restored if needed, see epmd documentation for details.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-8780</p> + </item> + <item> + <p> Building in a source tree without prebuilt platform + independent build results failed on the SSL examples + when: </p> <list><item> cross building. This has been + solved by not building the SSL examples during a cross + build. </item><item> building on Windows. </item></list> + <p> + Own Id: OTP-8791</p> + </item> + <item> + <p> + inet:getsockopt for SCTP sctp_default_send_param had a + bug to not initialize required feilds causing random + answers. It is now corrected.</p> + <p> + Own Id: OTP-8795 Aux Id: seq11655 </p> + </item> + <item> + <p> + The hipe_bifs:get_hrvtime/0 BIF now always returns a real + value even if the "perfctr" Linux kernel extension is not + available. It used to return a dummy value. (Thanks to + Mikael Pettersson.)</p> + <p> + Own Id: OTP-8798</p> + </item> + <item> + <p> + Calling a native-code compiled module with an + <c>on_load</c> function could cause a crash. (Thanks to + Mikael Pettersson.)</p> + <p> + Own Id: OTP-8799</p> + </item> + <item> + <p>The emulator could crash while writing a crash dump if + native-compiled modules had been loaded. (Thanks to Paul + Guyot.)</p> + <p> + Own Id: OTP-8801</p> + </item> + <item> + <p> + The garbage collector could crash if invoked from + native-compiled code after a call to a BIF. (Thanks to + Paul Guyot.)</p> + <p> + Own Id: OTP-8821</p> + </item> + <item> + <p> + A rare memory leak in binary:matches is removed</p> + <p> + Own Id: OTP-8823</p> + </item> + <item> + <p>For a socket in the HTTP packet mode, the return value + from <c>gen_tcp:recv/2,3</c> if there is an error in the + header will be <c>{ok,{http_error,String}}</c> instead of + <c>{error,{http_error,String}}</c> to be consistent with + <c>ssl:recv/2,3</c>.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-8831</p> + </item> + </list> + </section> + + + <section><title>Improvements and New Features</title> + <list> + <item> + <p> + ets:select_reverse/{1,2,3} are now documented.</p> + <p> + Own Id: OTP-7863</p> + </item> + <item> + <p> + External format of integers changed to make full use of + all 32 bits of INTEGER_EXT. This is a compatible change + as old code can read full 32-bit integers but only + produce 28-bit integers as INTEGER_EXT.</p> + <p> + Own Id: OTP-8540 Aux Id: seq11534 </p> + </item> + <item> + <p> + Large parts of the <c>ethread</c> library have been + rewritten. The <c>ethread</c> library is an Erlang + runtime system internal, portable thread library used by + the runtime system itself.</p> + <p> + Most notable improvement is a reader optimized rwlock + implementation which dramatically improve the performance + of read-lock/read-unlock operations on multi processor + systems by avoiding ping-ponging of the rwlock cache + lines. The reader optimized rwlock implementation is used + by miscellaneous rwlocks in the runtime system that are + known to be read-locked frequently, and can be enabled on + ETS tables by passing the <seealso + marker="stdlib:ets#new_2_read_concurrency">{read_concurrency, + true}</seealso> option upon table creation. See the + documentation of <seealso + marker="stdlib:ets#new/2">ets:new/2</seealso> for more + information. The reader optimized rwlock implementation + can be fine tuned when starting the runtime system. For + more information, see the documentation of the <seealso + marker="erts:erl#+rg">+rg</seealso> command line argument + of <c>erl</c>.</p> + <p> + There is also a new implementation of rwlocks that is not + optimized for readers. Both implementations interleaves + readers and writers during contention as opposed to, + e.g., the NPTL (Linux) pthread rwlock implementation + which use either a reader or writer preferred strategy. + The reader/writer preferred strategy is problematic since + it starves threads doing the non-preferred operation.</p> + <p> + The new rwlock implementations in general performs better + in ERTS than common pthread implementations. However, in + some extremely heavily contended cases this is not the + case. Such heavy contention can more or less only appear + on ETS tables. This when multiple processes do very large + amounts of write locked operations simultaneously on the + same table. Such use of ETS is bad regardless of rwlock + implementation, will never scale, and is something we + strongly advise against.</p> + <p> + The new rwlock implementations depend on atomic + operations. If no native atomic implementation is found, + a fallback solution will be used. Using the fallback + implies a performance degradation. That is, it is more + important now than before to build OTP with a native + atomic implementation.</p> + <p> + The <c>ethread</c> library contains native atomic + implementations for, x86 (32 and 64 bit), powerpc (32 + bit), sparc V9 (32 and 64 bit), and tilera (32 bit). On + other hardware gcc's builtin support for atomic memory + access will be used if such exists. If no such support is + found, <c>configure</c> will warn about no atomic + implementation available.</p> + <p> + The <c>ethread</c> library can now also use the + <c>libatomic_ops</c> library for atomic memory accesses. + This makes it possible for the Erlang runtime system to + utilize optimized native atomic operations on more + platforms than before. If <c>configure</c> warns about no + atomic implementation available, try using the + <c>libatomic_ops</c> library. Use the <seealso + marker="doc/installation_guide:INSTALL#How-to-Build-and-Install-ErlangOTP_A-Closer-Look-at-the-individual-Steps_Configuring">--with-libatomic_ops=PATH</seealso> + <c>configure</c> command line argument when specifying + where the <c>libatomic_ops</c> installation is located. + The <c>libatomic_ops</c> library can be downloaded from: + <url + href="http://www.hpl.hp.com/research/linux/atomic_ops/">http://www.hpl.hp.com/research/linux/atomic_ops/</url></p> + <p> + The changed API of the <c>ethread</c> library has also + caused modifications in the Erlang runtime system. + Preparations for the to come "delayed deallocation" + feature has also been done since it depends on the + <c>ethread</c> library.</p> + <p> + <em>Note</em>: When building for x86, the <c>ethread</c> + library will now use instructions that first appeared on + the pentium 4 processor. If you want the runtime system + to be compatible with older processors (back to 486) you + need to pass the <seealso + marker="doc/installation_guide:INSTALL#How-to-Build-and-Install-ErlangOTP_A-Closer-Look-at-the-individual-Steps_Configuring">--enable-ethread-pre-pentium4-compatibility</seealso> + <c>configure</c> command line argument when configuring + the system.</p> + <p> + Own Id: OTP-8544</p> + </item> + <item> + <p> + erlang:localtime_to_universaltime({{2008, 8, 1}, {0, 0, + 0}},true) when TZ=UTC now behaves consistently on all + Unix platforms.</p> + <p> + The problem fixed was originally reported by Paul Guyot + on erlang-bugs mailing list:</p> + <p> + http://www.erlang.org/pipermail/erlang-bugs/2008-November/001077.html</p> + <p> + Own Id: OTP-8580</p> + </item> + <item> + <p> + Optimization reducing memory consumption by two words per + ETS object.</p> + <p> + Own Id: OTP-8737</p> + </item> + <item> + <p> + Fixes for unsupported halfword-emulator</p> + <p> + Own Id: OTP-8745</p> + </item> + <item> + <p> + NIF 64-bit integer support; <c>enif_get_int64</c>, + <c>enif_get_uint64</c>, <c>enif_make_int64</c>, + <c>enif_make_uint64</c>.</p> + <p> + Own Id: OTP-8746</p> + </item> + <item> + <p> + Alignment of trailing data in messages has been adjusted. + This in order to be able to pass data of any type as + trailing data in the future.</p> + <p> + Own Id: OTP-8754</p> + </item> + <item> + <p> + The obsolete/driver.h header file has been removed. It + has been obsolete and deprecated since R8B. Drivers that + still include obsolete/driver.h must be updated to + include erl_driver.h.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-8758</p> + </item> + <item> + <p> + Added erlang:system_info(build_type) which makes it + easier to chose drivers, NIF libraries, etc based on + build type of the runtime system.</p> + <p> + The NIF library for crypto can now be built for valgrind + and/or debug as separate NIF libraries that will be + automatically loaded if the runtime system has been built + with a matching build type.</p> + <p> + Own Id: OTP-8760</p> + </item> + <item> + <p> + Further lessened the memory requirements of ETS objects.</p> + <p> + Own Id: OTP-8762</p> + </item> + <item> + <p>The broken elib_malloc alternate memory allocator has + been removed. <c>erlang:system_info(elib_malloc)</c> will + always return <c>false</c>, and in R15, + <c>erlang:system_info(elib_malloc)</c> will fail with a + <c>badarg</c> exception.</p> + <p> + Own Id: OTP-8764</p> + </item> + <item> + <p> + Calling <c>erlang:system_info/1</c> with the new argument + <c>update_cpu_info</c> will make the runtime system + reread and update the internally stored CPU information. + For more information see the documentation of <seealso + marker="erlang#update_cpu_info">erlang:system_info(update_cpu_info)</seealso>.</p> + <p> + The CPU topology is now automatically detected on Windows + systems with less than 33 logical processors. The runtime + system will now, also on Windows, by default bind + schedulers to logical processors using the + <c>default_bind</c> bind type if the amount of schedulers + is at least equal to the amount of logical processors + configured, binding of schedulers is supported, and a CPU + topology is available at startup.</p> + <p> + Own Id: OTP-8765</p> + </item> + <item> + <p> + The SMP ERTS internal child waiter thread used on Linux + system with NPTL was unintentionally disabled during + cross compilation rewrites (OTP-8323 in R13B03). It has + now been re-enabled. Enabling it again gives a slight + performance improvement.</p> + <p> + Own Id: OTP-8774</p> + </item> + <item> + <p> + <c>epmd</c> used to generate a message to the syslog when + it started up, which could be annoying. This has been + changed to only generate the message if the debug swith + is given. (Thanks to Michael Santos.)</p> + <p> + Own Id: OTP-8775</p> + </item> + <item> + <p> + The scheduler wakeup threshold is now possible to adjust + at system boot. For more information see the <seealso + marker="erl#+swt">+swt</seealso> command line argument of + <c>erl</c>.</p> + <p> + Own Id: OTP-8811</p> + </item> + <item> + <p> + The undocumented function inet:ifget/2 has been improved + to return interface hardware address (MAC) on platforms + supporting getaddrinfo() (such as BSD unixes). Note it + still does not work on all platforms for example not + Windows nor Solaris, so the function is still + undocumented.</p> + <p> + Buffer overflow and field init bugs for inet:ifget/2 and + inet:getservbyname/2 has also been fixed.</p> + <p> + Thanks to Michael Santos.</p> + <p> + Own Id: OTP-8816</p> + </item> + <item> + <p> + Optimizations for MIPS when using gcc atomics. (Thanks to + Steve Vinoski)</p> + <p> + Own Id: OTP-8834</p> + </item> + <item> + <p> + Lock optimization in timer functionality.</p> + <p> + Own Id: OTP-8835</p> + </item> + </list> + </section> + + + <section><title>Known Bugs and Problems</title> + <list> + <item> + <p>Fix epmd and build environment to build on VxWorks</p> + <p> + Own Id: OTP-8838</p> + </item> + </list> + </section> + +</section> + +<section><title>Erts 5.8</title> + + <section><title>Fixed Bugs and Malfunctions</title> + <list> + <item> + <p> + Driver threads, such as async threads, using <seealso + marker="erl_driver#ErlDrvPDL">port data locks</seealso> + peeked at the port status field without proper locking + when looking up the driver queue.</p> + <p> + Own Id: OTP-8475</p> + </item> + <item> + <p> + The use of <c>mmap()</c> was unnecessarily disabled when + cross compiling.</p> + <p> + The <c>configure</c> arguments <c>--with-ssl</c>, and + <c>--with-odbc</c> refused to accept libraries outside of + <c>$erl_xcomp_sysroot</c> when cross compiling for no + good reason.</p> + <p> + The <c>configure</c> argument <c>--with-odbc</c> didn't + handle the value <c>yes</c> correct.</p> + <p> + The <c>configure</c> arguments <c>--with-odbc</c>, and + <c>--without-odbc</c> have also been added to the + configure help.</p> + <p> + (Thanks to Steve Vinoski for reporting these issues)</p> + <p> + Own Id: OTP-8484</p> + </item> + <item> + <p> + A call to the BIF <c>unregister(RegName)</c> when a port + had the name <c>RegName</c> registered in the runtime + system without SMP support caused a runtime system crash. + (Thanks to Per Hedeland for the bugfix and test case.)</p> + <p> + Own Id: OTP-8487</p> + </item> + <item> + <p> + The runtime system crashed if fewer logical processors + were found than reported by <c>sysconf( + SC_NPROCESSORS_CONF)</c>.</p> + <p> + Own Id: OTP-8549</p> + </item> + <item> + <p> + Fix memory management bug causing crash of non-SMP + emulator with async threads enabled. The bug did first + appear in R13B03.</p> + <p> + Own Id: OTP-8591 Aux Id: seq11554 </p> + </item> + <item> + <p> + Port locks could be prematurely destroyed.</p> + <p> + Own Id: OTP-8612</p> + </item> + <item> + <p>The <c>empd</c> program could loop and consume 100% + CPU time if an unexpected error ocurred in + <c>listen()</c> or <c>accept()</c>. Now <c>epmd</c> will + terminate if a non-recoverable error occurs. (Thanks to + Michael Santos.)</p> + <p> + Own Id: OTP-8618</p> + </item> + <item> + <p> + When kernel poll has been enabled, a livelock could in + rare circumstances occur. Problem reported by Chetan + Ahuja, fix by Mikael Pettersson.</p> + <p> + Own Id: OTP-8632</p> + </item> + <item> + <p> + Windows: Closing port of program that stalled without + reading all data could deadlock scheduler thread.</p> + <p> + Own Id: OTP-8641</p> + </item> + <item> + <p> + On some combination of Montavista Linux on Cavium Octeon + processors, some socket-related system calls returned + other numbers than -1 for errors. This caused a core dump + in inet_drv.c. Now the code works around this problem.</p> + <p> + Own Id: OTP-8654</p> + </item> + <item> + <p> + Missing memory barriers in <c>erts_poll()</c> could cause + the runtime system to hang indefinitely.</p> + <p> + Own Id: OTP-8658</p> + </item> + <item> + <p> + <c>ethr_rwmutex_tryrlock()</c> acquired and refused to + acquire a lock with inverted logic. The lock was however + never acquired in a thread unsafe manner. (Thanks to JR + Zhang for noting this issue)</p> + <p> + Own Id: OTP-8662</p> + </item> + <item> + <p> + Extreme combinations of register/unregister in a highly + parallell SMP application could crash the VM. The error + is corrected.</p> + <p> + Own Id: OTP-8663</p> + </item> + <item> + <p> + On Windows, files are now opened with FILE_SHARE_DELETE + to get closer to Unix semantics.</p> + <p> + Own Id: OTP-8667</p> + </item> + <item> + <p> + <c>erlang:system_info(multi_scheduling)</c> sometimes + erroneously returned <c>enabled</c> when it should have + returned <c>blocked</c>.</p> + <p> + Own Id: OTP-8675</p> + </item> + <item> + <p> + Fix bug causing <c>erlang:decode_packet</c> and + <c>enif_make_string</c> to generate faulty strings with + negative character values for ascii values larger than + 127. (Thanks to Paul Guyot)</p> + <p> + Own Id: OTP-8685</p> + </item> + <item> + <p> + <c>open_port/2</c> with the <c>spawn</c> and + <c>spawn_executable</c> options can include an + <c>{env,Env}</c> option. In some cases unsetting + variables would not work on Unix (typically if more + variables were unset than were actually present in the + environment).</p> + <p> + Own Id: OTP-8701</p> + </item> + <item> + <p> + A user defined CPU topology set via a call to <seealso + marker="erlang#system_flag_cpu_topology">erlang:system_flag(cpu_topology, + CPUTopology)</seealso> was not properly verified, and + could in worst case cause an emulator crash. The emulator + crash could only occur when a user defined CPU topology + already existed and was redefined.</p> + <p> + Own Id: OTP-8710</p> + </item> + </list> + </section> + + + <section><title>Improvements and New Features</title> + <list> + <item> + <p> + The grammar for match specifications in ERTS users guide + only described the tracing dialect of match + specifications. An informal grammar for the ETS dialect + is added.</p> + <p> + Own Id: OTP-8086 Aux Id: seq11333 </p> + </item> + <item> + <p> + The module binary from EEP31 (and EEP9) is implemented.</p> + <p> + Own Id: OTP-8217</p> + </item> + <item> + <p> + New NIF API function <c>enif_make_new_binary</c></p> + <p> + Own Id: OTP-8474</p> + </item> + <item> + <p> + The guard BIF <c>is_boolean/1</c> (introduced in R10B) + has now been included in the lists of BIFs allowed in + guards in the Reference Manual.</p> + <p> + Own Id: OTP-8477</p> + </item> + <item> + <p> + Added function <c>zip:foldl/3</c> to iterate over zip + archives.</p> + <p> + Added functions to create and extract escripts. See + <c>escript:create/2</c> and <c>escript:extract/2</c>.</p> + <p> + The undocumented function <c>escript:foldl/3</c> has been + removed. The same functionality can be achieved with the + more flexible functions <c>escript:extract/2</c> and + <c>zip:foldl/3</c>.</p> + <p> + Record fields has been annotated with type info. Source + files as been adapted to fit within 80 chars and trailing + whitespace has been removed.</p> + <p> + Own Id: OTP-8521</p> + </item> + <item> + <p>A regular expression with many levels of parenthesis + could cause a buffer overflow. That has been corrected. + (Thanks to Michael Santos.)</p> + <p> + Own Id: OTP-8539</p> + </item> + <item> + <p> + <c>erlang:decode_packet(httph_bin,..)</c> could return + corrupt header strings or even crash the VM. This has + been fixed. It only happened on 32-bit VM if the header + name was unknown and between 16 and 20 characters long. + Sockets with simular <c>packet</c> option did not suffer + from this bug.</p> + <p> + Own Id: OTP-8548</p> + </item> + <item> + <p>New NIF features:</p> <list><item> Send messages from + a NIF, or from thread created by NIF, to any local + process (<c>enif_send</c>) </item><item> Store terms + between NIF calls (<c>enif_alloc_env</c>, + <c>enif_make_copy</c>) </item><item> Create binary terms + with user defined memory management + (<c>enif_make_resource_binary</c>) </item></list> <p>And + some incompatible changes made to the API. For more + information see the warning text in <seealso + marker="erl_nif">erl_nif(3)</seealso>.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-8555</p> + </item> + <item> + <p>If the '<c>fop</c>' program (needed for building PDF + files) cannot not be found, it is now possible to build + the HTML and man pages anyway (there will also be dummy + PDF files with no real content created). (Thanks to + Tuncer Ayaz.)</p> + <p> + Own Id: OTP-8559</p> + </item> + <item> + <p>When defining macros the closing right parenthesis + before the dot is now mandatory.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-8562</p> + </item> + <item> + <p>Local and imported functions now override the + auto-imported BIFs when the names clash. The pre R14 + behaviour was that auto-imported BIFs would override + local functions. To avoid that old programs change + behaviour, the following will generate an error:</p> + <list><item><p>Doing a call without explicit module name + to a local function having a name clashing with the name + of an auto-imported BIF that was present (and + auto-imported) before OTP R14A</p></item> + <item><p>Explicitly importing a function having a name + clashing with the name of an autoimported BIF that was + present (and autoimported) before OTP R14A</p></item> + <item><p>Using any form of the old compiler directive + <c>nowarn_bif_clash</c></p></item> </list> <p>If the BIF + was added or auto-imported in OTP R14A or later, + overriding it with an import or a local function will + only result in a warning,</p> <p>To resolve clashes, you + can either use the explicit module name <c>erlang</c> to + call the BIF, or you can remove the auto-import of that + specific BIF by using the new compiler directive + <c>-compile({no_auto_import,[F/A]}).</c>, which makes all + calls to the local or imported function without explicit + module name pass without warnings or errors.</p> <p>The + change makes it possible to add auto-imported BIFs + without breaking or silently changing old code in the + future. However some current code ingeniously utilizing + the old behaviour or the <c>nowarn_bif_clash</c> compiler + directive, might need changing to be accepted by the + compiler.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-8579</p> + </item> + <item> + <p> + A bug in re that could cause certain regular expression + matches never to terminate is corrected. (Thanks to + Michael Santos and Gordon Guthrie.)</p> + <p> + Own Id: OTP-8589</p> + </item> + <item> + <p>The erlang:open_port spawn and spawn_executable + directives can include an <c>{env,Env}</c> directive to + set up environment variables for the spawned process. A + bug prevented applications from using <c>{env,Env}</c> to + set an environment variable whose value ended with a + '<c>=</c>' (equal sign) character; the trailing equal + sign was mistaken as an indication that an environment + variable was to be cleared from the environment of the + spawned process. (Thanks to Steve Vinoski.)</p> + <p> + Own Id: OTP-8614</p> + </item> + <item> + <p><c>receive</c> statements that can only read out a + newly created reference are now specially optimized so + that it will execute in constant time regardless of the + number of messages in the receive queue for the process. + That optimization will benefit calls to + <c>gen_server:call()</c>. (See <c>gen:do_call/4</c> for + an example of a receive statement that will be + optimized.)</p> + <p> + Own Id: OTP-8623</p> + </item> + <item> + <p> + The functions file:advise/4 and file:datasync/1 have been + added. (Thanks to Filipe David Manana.)</p> + <p> + Own Id: OTP-8637</p> + </item> + <item> + <p> + New NIF API functions: <c>enif_make_atom_len</c>, + <c>enif_make_existing_atom_len</c>, + <c>enif_make_string_len</c>, <c>enif_get_atom_length</c>, + <c>enif_get_list_length</c>, <c>enif_is_list</c>, + <c>enif_is_tuple</c> (by Tuncer Ayaz)</p> + <p> + Own Id: OTP-8640</p> + </item> + <item> + <p> + Support for using gcc's built-in functions for atomic + memory access has been added. This functionallity will be + used if available and no other native atomic + implementation in ERTS is available.</p> + <p> + Own Id: OTP-8659</p> + </item> + <item> + <p> + The number of spinlocks used when implementing atomic + fall-backs when no native atomic implementation is + available has been increased from 16 to 1024.</p> + <p> + Own Id: OTP-8660</p> + </item> + <item> + <p> + Writer preferred pthread read/write locks has been + enabled on Linux.</p> + <p> + Own Id: OTP-8661</p> + </item> + <item> + <p> + The runtime system will by default bind schedulers to + logical processors using the <c>default_bind</c> bind + type if the amount of schedulers are at least equal to + the amount of logical processors configured, binding of + schedulers is supported, and a CPU topology is available + at startup. </p> + <p> + <em>NOTE:</em> 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. If this is the case you, are + are advised to unbind the schedulers using the <seealso + marker="erl#+sbt">+sbtu</seealso> command line argument, + or by invoking <seealso + marker="erlang#system_flag_scheduler_bind_type">erlang:system_flag(scheduler_bind_type, + unbound)</seealso>.</p> + <p> + Own Id: OTP-8666</p> + </item> + <item> + <p> + The recently added BIFs erlang:min/2, erlang:max/2 and + erlang:port_command/3 are now auto-imported (as they were + originally intended to be). Due to the recent compiler + change (OTP-8579), the only impact on old code defining + it's own min/2, max/2 or port_command/3 functions will be + a warning, the local functions will still be used. The + warning can be removed by using + -compile({no_auto_import,[min/2,max/2,port_command/3]}). + in the source file.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-8669 Aux Id: OTP-8579 </p> + </item> + <item> + <p> + There is a new option 'exclusive' to file:open/2 that + uses the OS O_EXCL flag where supported to open the file + in exclusive mode.</p> + <p> + Own Id: OTP-8670</p> + </item> + <item> + <p> + Now, binary_to_term/2 is auto-imported. This will cause a + compile warning if and only if a module has got a local + function with that name.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-8671</p> + </item> + <item> + <p> + Alignment of scheduler data and run queues were adjusted.</p> + <p> + Own Id: OTP-8673</p> + </item> + <item> + <p>Call time breakpoint tracing</p> <list><item>Introduce + a <c>call_time</c> option to + <c>erlang:trace_pattern/3</c>.This option enables call + time breakpoint tracing on code that is executed by + processes with call tracing enabled. Call time tracing + stores the number of calls and the time spent of each + function with this trace pattern enabled. The information + can be retrieved with <c>erlang:trace_info/2</c></item> + <item>Add a scheduler array for BpData. To solve the + issue of multiple schedulers constantly updating the head + pointer to the bp data wheel, each scheduler now has its + own entrypoint to the wheel. This head pointer can be + updated without a locking being taken.</item></list> + <p>Teach call count tracing to use atomics</p> + <list><item>Call count previously used a global lock for + accessing and writing its counter in the breakpoint. This + is now changed to atomics instead.</item> <item>The + change will let call count tracing and cprof to scale + better when increasing the number of + schedulers.</item></list> + <p> + Own Id: OTP-8677</p> + </item> + <item> + <p><c>eprof</c> has been reimplemented with support in + the Erlang virtual machine and is now both faster (i.e. + slows down the code being measured less) and scales much + better. In measurements we saw speed-ups compared to the + old eprof ranging from 6 times (for sequential code that + only uses one scheduler/core) up to 84 times (for + parallel code that uses 8 cores).</p> + <p>Note: The API for the <c>eprof</c> has been cleaned up + and extended. See the documentation.</p> + <p> + *** POTENTIAL INCOMPATIBILITY ***</p> + <p> + Own Id: OTP-8706</p> + </item> + </list> + </section> + +</section> + <section><title>Erts 5.7.5.1</title> <section><title>Fixed Bugs and Malfunctions</title> @@ -144,7 +1127,7 @@ <item> <p>A number of bugs concerning re and unicode are corrected:</p> - <p>re:compile no longer looses unicode option, which also + <p>re:compile no longer loses unicode option, which also fixes bug in re:split.</p> <p>re:replace now handles unicode charlist replacement argument</p> @@ -4716,7 +5699,7 @@ </item> <item> <p>The runtime system with SMP support did not slowly adjust - it's view of time when the system time suddenly changed.</p> + its view of time when the system time suddenly changed.</p> <p>Timeouts could sometimes timeout too early on the runtime system with SMP support.</p> <p>Own Id: OTP-6202</p> |