duktape

Commit Graph

Author	SHA1	Message	Date
Sami Vaarala	098667af51	Use duktape-releases repo for website build	10 years ago
Sami Vaarala	c1615ad65a	Remove Duktape binaries These won't be edited out of the repo, but no further binaries will be added.	10 years ago
Sami Vaarala	f18c5f9a50	Merge branch 'add-git-conventions-doc'	10 years ago
Sami Vaarala	d0d6a5cf9e	Add current (rough) Git conventions	10 years ago
Sami Vaarala	fae5bf13ae	Ditz issue update	10 years ago
Sami Vaarala	2a5adb0f56	Ditz issue update	10 years ago
Sami Vaarala	adfd9ddd76	Merge branch 'fix-gc-torture-issues'	10 years ago
Sami Vaarala	367c66ae2e	Release note: mark-and-sweep zero realloc()	10 years ago
Sami Vaarala	b51973b5b0	Clarify realloc() 0-size handling in a few places In these situations the new size being allocated can't be zero, so the call site can just compare the result pointer and assume NULL means error. It's not an obvious thing so add an explicit comment and an assert.	10 years ago
Sami Vaarala	966c0fa511	Realloc fixes to heap memory handling Discovered with GC torture testing: when mark-and-sweep is triggered during a realloc, the realloc() retry handler does not handle zero size correctly (i.e. it doesn't accept a NULL as a successful result in that case). After these fixes GC torture tests pass.	10 years ago
Sami Vaarala	a9e98e9a0a	Ditz issue update	10 years ago
Sami Vaarala	061b2efa52	Merge branch 'guide-limitations-luacomparison-cleanups'	10 years ago
Sami Vaarala	f2a1502cf7	Weak references to Lua comparison	10 years ago
Sami Vaarala	35149f42ea	Remove finalizers from Lua comparison section Lua 5.1 does have finalizers in the form of the __gc metamethod (which only works for userdata objects).	10 years ago
Sami Vaarala	3511614127	Remove dollar escape as RegExp limitation This feature is intentional and there's actually need for more real world compatible behavior for RegExps.	10 years ago
Sami Vaarala	cbf8f6173d	Merge branch 'guide-limitations-threading-link'	10 years ago
Sami Vaarala	e76acc2827	Threading link to limitations	10 years ago
Sami Vaarala	d604107586	Merge branch 'initjs-comment-trivia'	10 years ago
Sami Vaarala	b73ae90aa2	Trivial comment fix	10 years ago
Sami Vaarala	ab7458da4d	Merge branch 'known-issues-test262-etc'	10 years ago
Sami Vaarala	281af80643	Add raw testcase runs to 1.0 known issues	10 years ago
Sami Vaarala	835ea1b679	Merge branch 'update-release-checklist'	10 years ago
Sami Vaarala	dae6fdc63e	Add matfix test to release checklist	10 years ago
Sami Vaarala	739eb47da3	Ditz issue update	10 years ago
Sami Vaarala	f45eb12323	Merge branch 'release-notes-v1.0.0' Add (draft) release notes for the 1.0 release. The intent is for the release notes of a stable version to be up-to-date so one can see known issues and workarounds at a glance.	10 years ago
Sami Vaarala	035fc7431e	Release notes 1.0.0 draft	10 years ago
Sami Vaarala	34983f87b7	Merge branch 'matrix-compile-test' Quick hack test for a mstrix compile test. This rudimentary test revealed a few compile warning issues which are fixed separately. For proper coverage the test matrices will be very large, so a random sampling approach is probably the most useful to quickly go through a lot of variations. The test matrix could then include basically every compiler and feature option in isolation (creating billions of combinations) as well as every testcase Duktape has. For a given test matrix configuration one integer can be used to identify a combination and easily report and repeat failures.	10 years ago
Sami Vaarala	ec02cd0b02	Matrix compile hack test	10 years ago
Sami Vaarala	9e1a0390b6	Merge branch 'fix-compile-warnings' Fix some compile warnings mostly related to GCC -O4. The GCC -O4 result is not clean when assertions are enabled; the remaining warnings seem spurious and difficult to silence in a reasonable way.	10 years ago
Sami Vaarala	633d194476	Fix (mostly) gcc -O4 related compile warnings Unused variables in some option combinations. Some of these were caused by having -DDUK_OPT_DEBUG but no debug print options. The GCC -O4 array subscript warning seems spurious and is probably caused by GCC not being 100% sure 'count' is always valid. The call sites for duk__advance_chars() have a variety of argument values, including those computed using shifts etc. Add an explicit check for the valid range which silences the GCC error. The check also protects against regressions that would cause an invalid count to be used. With the check in place, potentially memory unsafe behavior will now cause an internal error instead. If the form DUK_ASSERT(x >= 0 && x < y) is used, gcc will cause a spurious warning: duk_api_stack.c:149:33: error: assuming signed overflow does not occur when assuming that (X - c) > X is always false [-Werror=strict-overflow] duk_error.h:131:8: note: in definition of macro ‘DUK_ASSERT’ By breaking down the assert into two, i.e. DUK_ASSERT(x >= 0) and DUK_ASSERT(x < y) the warning disappears.	10 years ago
Sami Vaarala	160fb86fca	Ditz issue update	10 years ago
Sami Vaarala	96f703fe91	Ditz issue update	10 years ago
Sami Vaarala	4af6161dfa	Merge branch 'fix-finalizer-skip-bug' Fix a mark-and-sweep bug where memory safety could be compromised when finalizer execution was disabled during a mark-and-sweep run.	10 years ago
Sami Vaarala	15e9389cba	Fix mark-and-sweep finalizer memory safety bug When one mark-and-sweep round leaves objects-to-be-finalized on the heap finalize_list, the next round must treat them like reachability roots to ensure any objects they point to get marked reachable. The REACHABLE flag was cleared for objects-to-be-finalized when they were moved from the heap_allocated list to the finalize_list. This is correct if finalizers are executed afterwards. But when finalizers are skipped, the next mark-and-sweep will mark the objects on the finalize_list REACHABLE which broke some assertions. Fix that by clearing REACHABLE flags for objects on the finalize_list explicitly after sweeping. This overlaps with the existing code in the sweep phase in a harmless way. This solution is a minimal fix to the issue. In Duktape 1.1 the intent is to rework mark-and-sweep so that finalizer execution can be independent of mark-and-sweep, so a root canal fix wouldn't be useful at this point.	10 years ago
Sami Vaarala	1a974e4dc5	Release note: mark-and-sweep finalizer skip	10 years ago
Sami Vaarala	08294a80d7	Testcase for skipped finalizers	10 years ago
Sami Vaarala	0f72534a8d	Ditz issue update	10 years ago
Sami Vaarala	da5aaa0bb4	Merge branch 'api-get-prop-rc'	10 years ago
Sami Vaarala	d662a1231c	Clarify non-zero return value is 1	10 years ago
Sami Vaarala	8cfca773c6	Ditz issue update	10 years ago
Sami Vaarala	9cca56f2cd	Merge branch 'fix-ecmatest-target'	10 years ago
Sami Vaarala	2fdacba7a9	Report diffs to other engines in make ecmatest	10 years ago
Sami Vaarala	a3cc463f4e	Add missing engine options for 'make ecmatest'	10 years ago
Sami Vaarala	e12ce70343	Merge branch 'add-robots-txt'	10 years ago
Sami Vaarala	8374b44704	Add robots.txt to site	10 years ago
Sami Vaarala	5ea33e01bf	Ditz issue update	10 years ago
Sami Vaarala	8cd9ce737f	Ditz issue update	10 years ago
Sami Vaarala	689029422e	Merge branch 'c-module-example' Add an example of a C module into the guide, even if there is not strongly recommended module convention. These conventions will be worked out probably for the 1.1 release, together with a suggested DLL format for major platforms. Together, these will allow modules to be more easily exchanged between code bases as pre-packaged binaries and such.	10 years ago
Sami Vaarala	85d4d6ed85	API testcase for C module example	10 years ago
Sami Vaarala	bc2b5a874b	Writing C modules placeholder section in guide For now, there is no fixed convention but provide some module anatomy example anyway.	10 years ago

1 2 3 4 5 ...

4115 Commits (a0b06df09204843620e229cae3b9e7dbe8944414) All Branches Search

4115 Commits (a0b06df09204843620e229cae3b9e7dbe8944414)

All Branches