kotovalexarian-likes-github/mini_racer
1
0
Fork 0
mirror of https://github.com/rubyjs/mini_racer synced 2023-03-27 23:21:28 -04:00
Code Releases Activity
mini_racer/CHANGELOG

226 lines
4.7 KiB
Text
Raw Normal View History

FEATURE: be specific about the version of libv8 This avoid issues going forward. libv8 tends to break against the wrong version of mini_racer closes #168
2020-07-22 18:28:20 -04:00
- 23-07-202
- 0.3.1
- FIX: specify that libv8 must be larger than 8.4.255 but smaller than 8.5, this avoids issues going forward
DEV: Changelog and version bump
2020-07-21 19:07:00 -04:00
- 22-07-2020
- 0.3.0
- FEATURE: upgraded to libv8 version 8.4.255.0
new release
2020-06-29 00:51:37 -04:00
- 29-06-2020
- 0.2.15
- FEATURE: basic wasm support via pump_message_loop
FEATURE: add support for Isolate#low_memory_notification This is another mechanism we can use for releasing as much memory as possible Note: Isolate#idle_notification can be used as well, but it is much slower to act. low_memory_notification forces a full GC and will clear up large amounts of space from the V8 heap. This also adds support for `ensure_gc_after_idle` options for MiniRacer::Context this allows you to automatically conserve memory on contexts, and only runs the GC when the context was idle for a certain amount of time.
2020-05-14 22:05:14 -04:00
- 15-05-2020
FIX: to keep api consistent ensure_gc_after_idle should be in ms Also introduce strict param checking for all params passed to context This avoids typo errors being totally hidden from user
2020-05-14 23:36:43 -04:00
- 0.2.14
- FIX: ensure_gc_after_idle should take in milliseconds like the rest of the APIs not seconds
- FEATURE: strict params on MiniRacer::Context.new
- 15-05-2020
FIX: edge case where ensure_gc_after_idle could fire when not idle We were missing checks to see if context was active
2020-05-14 22:57:26 -04:00
- 0.2.13
FIX: to keep api consistent ensure_gc_after_idle should be in ms Also introduce strict param checking for all params passed to context This avoids typo errors being totally hidden from user
2020-05-14 23:36:43 -04:00
FIX: edge case where ensure_gc_after_idle could fire when not idle We were missing checks to see if context was active
2020-05-14 22:57:26 -04:00
- FIX: edge case around ensure_gc_after_idle possibly firing when context is not idle
- 15-05-2020
FEATURE: add support for Isolate#low_memory_notification This is another mechanism we can use for releasing as much memory as possible Note: Isolate#idle_notification can be used as well, but it is much slower to act. low_memory_notification forces a full GC and will clear up large amounts of space from the V8 heap. This also adds support for `ensure_gc_after_idle` options for MiniRacer::Context this allows you to automatically conserve memory on contexts, and only runs the GC when the context was idle for a certain amount of time.
2020-05-14 22:05:14 -04:00
- 0.2.12
- FEATURE: isolate.low_memory_notification which can force a full GC
- FEATURE: MiniRacer::Context.new(ensure_gc_after_idle: 2) - to force full GC 2 seconds after context is idle, this allows you to conserve memory on isolates
Bump version of mini_racer
2020-05-14 03:24:02 -04:00
- 14-05-2020
- 0.2.11
- FIX: dumping heap snapshots was not flushing the file leading to corrupt snapshots
- FIX: a use-after-free shutdown crash
update changelog
2020-04-21 22:10:15 -04:00
- 0.2.10
- 22-04-2020
- FEATURE: memory softlimit support for nogvl_context_call
bump version
2020-01-08 20:25:02 -05:00
- 0.2.9
- 09-01-2020
- FIX: correct segfault when JS returns a Symbol and properly cast to ruby symbol
FIX: ensure thread is always stopped on timeout This ensures that even if we see an exception we will take care of stopping the timeout thread. Previously the race condition left an edge case where thread would terminate after pipe was closed
2019-11-10 20:52:44 -05:00
- 0.2.8
- 11-11-2019
- FIX: ensure thread live cycle is properly accounter for following file descriptor fix
FIX: ensure we release pipe fds early Previously we relied on GC to release pipe fds. This worked on earlier versions of Ruby but on 2.7 it will start breaking. We now release the fds early to avoid the issues.
2019-11-10 20:20:26 -05:00
- 0.2.7
- 11-11-2019
- FIX: release the file descriptor for timeout pipe earlier (this avoids holding too many files open in Ruby 2.7)
bump version of mini racer
2019-05-14 03:57:41 -04:00
- 14-05-2019
- 0.2.6
- FEATURE: add support for write_heap_snapshot which helps you analyze memory
Release new version This is urgent, cause libv8 no longer works with currently released mini_racer.
2019-04-25 01:15:25 -04:00
- 25-04-2019
- 0.2.5
Introduce compatibility with V8 7.3 (#138) * Mechanical removal of use of non-maybe String::ToObject() Due to the non-maybe version of String::ToObject() being deprecated and altogether removed from V8 [1] it is necessary to migrate to using the maybe version. This commit is a mechanical change that uses the context at hand when calling String::ToObject() to pass it to it. The resulting MaybeLocal is then unwrapped with MaybeLocal::ToLocalChecked() as I consider the verifications performed on the String instances to be sufficient to ensure no crashes. [1] https://chromium-review.googlesource.com/c/v8/v8/+/1172350/ https://chromium.googlesource.com/v8/v8/+/c8376b0069ebe16c67acf90c3cda3457ddccba4f * Mechanical removal of use of non-maybe Local<T>::ToString() Due to the non-maybe version of Local<T>::ToString() being deprecated and altogether removed from V8 [1] it is necessary to migrate to using the maybe version. This commit is a mechanical change that uses the context at hand when calling Local<T>::ToString() to pass it to it. The resulting MaybeLocal is then unwrapped with MaybeLocal::ToLocalChecked() as I consider the context of the uses to be sufficiently safe. [1] https://chromium-review.googlesource.com/c/v8/v8/+/1172350/ https://chromium.googlesource.com/v8/v8/+/c8376b0069ebe16c67acf90c3cda3457ddccba4f * Mechanical removal of the use of non-maybe Value::Int32Value() and NumberValue() Due to the non-maybe version of Value::Int32Value() and Value::NumberValue() being deprecated and altogether removed from V8 [1] it is necessary to migrate to using the maybe versions. This commit is a mechanical change that uses the context at hand when calling Value::Int32Value() or Value::NumberValue() to pass it to the respective function. The resulting Maybe is then unwrapped with Maybe::ToChecked() as I consider the verifications performed on the Value instances to be sufficient to ensure no crashes. [1] https://chromium-review.googlesource.com/c/v8/v8/+/1172350/ https://chromium.googlesource.com/v8/v8/+/c8376b0069ebe16c67acf90c3cda3457ddccba4f * Mechanical removal of use of String::Utf8Length() without isolate Due to the version of String::Utf8Length() with no paramters being deprecated and altogether removed from V8 [1] it is necessary to migrate to using the version that accepts isolate as a parameter. This commit is a mechanical change that uses the isolate available in the context where String::Utf8Length() is called. This is a breaking change imposing use of V8 6.9.411 or up. [1] https://chromium-review.googlesource.com/c/v8/v8/+/1124724/ https://chromium.googlesource.com/v8/v8/+/3dd5c6fe38355b8323597341409b37f931de5a85 * Remove the uses of deprecated snapshot-related functions Due to V8::CreateSnapshotDataBlob() and V8::WarmUpSnapshotDataBlob() being deprecated and altogether removed from V8 [1] it is necessary to migrate to using local implementations of them. This commit introduces create_snapshot_data_blob(), warm_up_snapshot_data_blob() and the helper function run_extra_code(). Their implementations have been copied over from [2]. [1] https://github.com/v8/v8/commit/b3738e658345adabaa958b9f9a94ca01fc87d5e4 https://chromium-review.googlesource.com/c/v8/v8/+/1019442/ [2] https://github.com/v8/v8/blob/7.3.492.27/test/cctest/test-serialize.cc https://chromium.googlesource.com/v8/v8.git/+/30602560a8fdb0bbfb50d70be867f32b72758a2f/test/cctest/test-serialize.cc * Non-trivial removal of uses of non-maybe Date::New() Due to the non-maybe version of Date::New() being deprecated and altogether removed from V8 [1] it is necessary to migrate to using the maybe version. This commit introduces a context argument to the convert_ruby_to_v8() function so it can be passed to the maybe version of Date::New(). This imposed changes throughout the code base so that the context can be passed together with the isolate to convert_ruby_to_v8(). [1] https://chromium-review.googlesource.com/c/v8/v8/+/1357056 https://github.com/v8/v8/commit/e84b92d7658dee79c34304a67c4fb84c93cd071b * Non-trivial removal of use of non-maybe Local<T>::ToString() This commit builds upon fe62f7935582bd889742ec77ee2289a8f6cb16e6. The use of Local<T>::ToString() in convert_result_to_ruby() did not have an immediately available context to use. This commit unwraps the context from p_ctx and passes it to the function and then unwraps the MaybeLocal result with MaybeLocal::ToLocalChecked() assuming the verification beforehand is sufficient to ensure there won't be any crashes. * Update the changelog * Replace placeholder in LICENSE.txt
2019-04-24 19:39:56 -04:00
- FIX: Compatiblity fixes for V8 7 and above @ignisf
- FIX: Memory leak in gc_callback @messense
- IMPROVEMENT: Added example of sourcemap support @ianks
Release new version This is urgent, cause libv8 no longer works with currently released mini_racer.
2019-04-25 01:15:25 -04:00
- URGENT: you will need this release for latest version of libv8 to work
Introduce compatibility with V8 7.3 (#138) * Mechanical removal of use of non-maybe String::ToObject() Due to the non-maybe version of String::ToObject() being deprecated and altogether removed from V8 [1] it is necessary to migrate to using the maybe version. This commit is a mechanical change that uses the context at hand when calling String::ToObject() to pass it to it. The resulting MaybeLocal is then unwrapped with MaybeLocal::ToLocalChecked() as I consider the verifications performed on the String instances to be sufficient to ensure no crashes. [1] https://chromium-review.googlesource.com/c/v8/v8/+/1172350/ https://chromium.googlesource.com/v8/v8/+/c8376b0069ebe16c67acf90c3cda3457ddccba4f * Mechanical removal of use of non-maybe Local<T>::ToString() Due to the non-maybe version of Local<T>::ToString() being deprecated and altogether removed from V8 [1] it is necessary to migrate to using the maybe version. This commit is a mechanical change that uses the context at hand when calling Local<T>::ToString() to pass it to it. The resulting MaybeLocal is then unwrapped with MaybeLocal::ToLocalChecked() as I consider the context of the uses to be sufficiently safe. [1] https://chromium-review.googlesource.com/c/v8/v8/+/1172350/ https://chromium.googlesource.com/v8/v8/+/c8376b0069ebe16c67acf90c3cda3457ddccba4f * Mechanical removal of the use of non-maybe Value::Int32Value() and NumberValue() Due to the non-maybe version of Value::Int32Value() and Value::NumberValue() being deprecated and altogether removed from V8 [1] it is necessary to migrate to using the maybe versions. This commit is a mechanical change that uses the context at hand when calling Value::Int32Value() or Value::NumberValue() to pass it to the respective function. The resulting Maybe is then unwrapped with Maybe::ToChecked() as I consider the verifications performed on the Value instances to be sufficient to ensure no crashes. [1] https://chromium-review.googlesource.com/c/v8/v8/+/1172350/ https://chromium.googlesource.com/v8/v8/+/c8376b0069ebe16c67acf90c3cda3457ddccba4f * Mechanical removal of use of String::Utf8Length() without isolate Due to the version of String::Utf8Length() with no paramters being deprecated and altogether removed from V8 [1] it is necessary to migrate to using the version that accepts isolate as a parameter. This commit is a mechanical change that uses the isolate available in the context where String::Utf8Length() is called. This is a breaking change imposing use of V8 6.9.411 or up. [1] https://chromium-review.googlesource.com/c/v8/v8/+/1124724/ https://chromium.googlesource.com/v8/v8/+/3dd5c6fe38355b8323597341409b37f931de5a85 * Remove the uses of deprecated snapshot-related functions Due to V8::CreateSnapshotDataBlob() and V8::WarmUpSnapshotDataBlob() being deprecated and altogether removed from V8 [1] it is necessary to migrate to using local implementations of them. This commit introduces create_snapshot_data_blob(), warm_up_snapshot_data_blob() and the helper function run_extra_code(). Their implementations have been copied over from [2]. [1] https://github.com/v8/v8/commit/b3738e658345adabaa958b9f9a94ca01fc87d5e4 https://chromium-review.googlesource.com/c/v8/v8/+/1019442/ [2] https://github.com/v8/v8/blob/7.3.492.27/test/cctest/test-serialize.cc https://chromium.googlesource.com/v8/v8.git/+/30602560a8fdb0bbfb50d70be867f32b72758a2f/test/cctest/test-serialize.cc * Non-trivial removal of uses of non-maybe Date::New() Due to the non-maybe version of Date::New() being deprecated and altogether removed from V8 [1] it is necessary to migrate to using the maybe version. This commit introduces a context argument to the convert_ruby_to_v8() function so it can be passed to the maybe version of Date::New(). This imposed changes throughout the code base so that the context can be passed together with the isolate to convert_ruby_to_v8(). [1] https://chromium-review.googlesource.com/c/v8/v8/+/1357056 https://github.com/v8/v8/commit/e84b92d7658dee79c34304a67c4fb84c93cd071b * Non-trivial removal of use of non-maybe Local<T>::ToString() This commit builds upon fe62f7935582bd889742ec77ee2289a8f6cb16e6. The use of Local<T>::ToString() in convert_result_to_ruby() did not have an immediately available context to use. This commit unwraps the context from p_ctx and passes it to the function and then unwraps the MaybeLocal result with MaybeLocal::ToLocalChecked() assuming the verification beforehand is sufficient to ensure there won't be any crashes. * Update the changelog * Replace placeholder in LICENSE.txt
2019-04-24 19:39:56 -04:00
version bump
2018-11-01 20:26:36 -04:00
- 02-11-2018
- 0.2.4
- FIX: deadlock releasing context when shared isolates are used
- FEATURE: include js backtrace when snapshots do not compile
Bump new version
2018-09-27 21:38:30 -04:00
- 28-09-2018
Drop condition to use libc++ on macOS According to Apple, projects should be migrated to libc++[1]: > Building with libstdc++ was deprecated with Xcode 8 and is not supported in > Xcode 10 when targeting iOS. C++ projects must now migrate to libc++ and are > recommended to set a deployment target of macOS 10.9 or later, or iOS 7 or > later. Besides changing the C++ Standard Library build setting, developers > should audit hard-coded linker flags and target dependencies to remove > references to libstdc++ (including -lstdc++, -lstdc++.6.0.9, > libstdc++.6.0.9.tbd, and libstdc++.6.0.9.dylib). Project dependencies such as > static archives that were built against libstdc++ will also need to be rebuilt > against libc++. (40885260) [1] https://developer.apple.com/documentation/xcode_release_notes/xcode_10_release_notes
2018-09-27 10:54:15 -04:00
Bump new version
2018-09-27 21:38:30 -04:00
- 0.2.3
- Drop all conditional logic from Mini Racer compilation for clang, always
rely on MacOS being High Sierra or up
Drop condition to use libc++ on macOS According to Apple, projects should be migrated to libc++[1]: > Building with libstdc++ was deprecated with Xcode 8 and is not supported in > Xcode 10 when targeting iOS. C++ projects must now migrate to libc++ and are > recommended to set a deployment target of macOS 10.9 or later, or iOS 7 or > later. Besides changing the C++ Standard Library build setting, developers > should audit hard-coded linker flags and target dependencies to remove > references to libstdc++ (including -lstdc++, -lstdc++.6.0.9, > libstdc++.6.0.9.tbd, and libstdc++.6.0.9.dylib). Project dependencies such as > static archives that were built against libstdc++ will also need to be rebuilt > against libc++. (40885260) [1] https://developer.apple.com/documentation/xcode_release_notes/xcode_10_release_notes
2018-09-27 10:54:15 -04:00
version bump
2018-09-25 19:37:14 -04:00
- 26-09-2018
- 0.2.2
- WORKAROUND: RUBY_PLATFORM is hardcoded on Ruby compile and can not be
trusted for feature detection, use a different technique when checking for
macOS Mojave
- 25-09-2018
Introduce changes to account for upstream V8 API deprecations
2018-06-21 11:06:16 -04:00
version bump
2018-09-24 19:49:42 -04:00
- 0.2.1
- FEATURE: Mojave macOS support
version bump
2018-09-25 19:37:14 -04:00
- 06-07-2018
version bump
2018-07-05 20:49:23 -04:00
- 0.2.0
FEATURE: we explicitly only support Ruby 2.3 and up - remove 2.2 from travis - be explicit in gemspec - add note to readme
2018-06-26 19:35:22 -04:00
- FEATURE: context#call to allow for cheaper invocation of functions
- FIX: rare memory leak when terminating a long running attached function
- FIX: rare segfault when terminating a long running attached function
- FIX: Reimplement Isolate#idle_notification using idle_notification_deadline, API remains the same @ignisf
Introduce changes to account for upstream V8 API deprecations
2018-06-21 11:06:16 -04:00
- Account for changes in the upstream V8 API @ignisf
version bump
2018-07-05 20:49:23 -04:00
- Support for libv8 6.7
Bump version closes #65
2017-12-27 16:52:56 -05:00
23-08-2017
- 0.1.15
- bump dependency of libv8 to 6.3
version bump
2017-08-23 12:17:33 -04:00
23-08-2017
correct dependency and do a version bump
2017-08-23 21:44:43 -04:00
- 0.1.14
- libv8 erronuously bumped to beta, reverted change
23-08-2017
version bump
2017-08-23 12:17:33 -04:00
- 0.1.13
- Fix: amend array buffer allocator to use v8 6.0 compatible allocator @ignisf
FEATURE: upgrade libv8 to 5.7
2017-07-18 10:54:23 -04:00
18-07-2017
bump version
2017-07-26 10:53:26 -04:00
- 0.1.12
- Feature: upgrade libv8 to 5.9
- Fix: warning when runnin with ruby warnings enabled (missed @disposed initialize)
18-07-2017
FEATURE: upgrade libv8 to 5.7
2017-07-18 10:54:23 -04:00
- 0.1.11
- Feature: upgrade libv8 to 5.7
Memory profiling and control features - #heap_stats for v8 heap stats - #dispose to quickly dispose of heap - fix minor memory leak
2017-07-13 17:43:43 -04:00
13-07-2017
- 0.1.10
FEATURE: filename support for #eval also fixes heap_stats so it returns all 0s when context is disposed
2017-07-17 11:05:33 -04:00
- Fix leak: memory leak when disposing a context (20 bytes per context)
Memory profiling and control features - #heap_stats for v8 heap stats - #dispose to quickly dispose of heap - fix minor memory leak
2017-07-13 17:43:43 -04:00
- Feature: added #heap_stats so you can get visibility from context to actual memory usage of isolate
- Feature: added #dispose so you reclaim all v8 memory right away as opposed to waiting for GC
FEATURE: filename support for #eval also fixes heap_stats so it returns all 0s when context is disposed
2017-07-17 11:05:33 -04:00
- Feature: you can now specify filename in an eval eg: eval('a = 1', filename: 'my_awesome.js')
Memory profiling and control features - #heap_stats for v8 heap stats - #dispose to quickly dispose of heap - fix minor memory leak
2017-07-13 17:43:43 -04:00
PERF: on boundary convert to and from json Previously we would walk the object graph recursively and convert, this proves to be extremely slow for large objects due to recursion and huge amount of v8 contexts that were created By converting objects to json on the boundary we also correct timeout semantics and so on.
2017-03-09 16:03:13 -05:00
09-03-2017
- 0.1.9
- Perf: speed up ruby/node boundary performance when moving large objects
bump version
2017-02-06 09:10:24 -05:00
06-02-2017
- 0.1.8
Update the CHANGELOG
2017-02-04 17:04:13 -05:00
- Fix: Include math.h to fix use of undeclared identifier floor with rbx. See #51
FIX: don't try to release forked isolates, just leak them
2016-11-02 07:18:00 -04:00
02-11-2016
- 0.1.7
- Fix: if for some reason an isolate was forked don't free it and raise a warning instead to avoid hanging process
FIX: timeout mechanics were not correct This fix ensures timeout will always be processed correctly, it also adds a safeguard to ensure against race conditions
2016-10-24 22:15:03 -04:00
25-10-2016
- 0.1.6
- Fix: timeout behavior was incorrect, in some cases stop could be called on already stopped contexts
Update CHANGELOG
2016-10-10 00:00:14 -04:00
10-10-2016
Bump version - depend on latest v8 - comment out test that is rather slow - add changelog
2016-10-09 17:43:18 -04:00
- 0.1.5
- Support for snapshots, shared isolates, runtime flags thanks to @wk8
- Fix timeout behavior when it occurs in an attached Ruby method
bump version add changelog
2016-05-19 08:15:54 -04:00
19-05-2016
version bump
2016-06-14 21:23:41 -04:00
- 0.1.4
- Set upper bound for libv8 inclusion @ignisf
- Support conversion of Date, Time and DateTime from Ruby to JS @seanmakesgames
- Support conversion of large numbers back from Ruby to JS @seanmakesgames
bump version add changelog
2016-05-19 08:15:54 -04:00
- 0.1.3
- Support more conversions from Ruby back to JS (Hash, Symbol, Array)
- Support attaching nested objects
add change log
2016-05-16 20:49:48 -04:00
17-05-2016
bump version add changelog
2016-05-19 08:15:54 -04:00
- 0.1.2
add change log
2016-05-16 20:49:48 -04:00
- Gemspec specifies minimal version of Ruby (2.0)
- Implement #load on Context to load files
bump version
2016-05-16 17:31:41 -04:00
17-05-2016
bump version add changelog
2016-05-19 08:15:54 -04:00
- 0.1.1
bump version
2016-05-16 17:31:41 -04:00
- Added unblock function so SIGINT does not lead to a crash
changelog
2016-05-14 03:11:40 -04:00
14-05-2016
bump version add changelog
2016-05-19 08:15:54 -04:00
- 0.1.1.beta.1
changelog
2016-05-14 03:11:40 -04:00
- First release
Copy permalink