Ryan VanderMeulen
a876ac1f03
Patches: - bug1410214.patch: regenerated for v1.7.3 filemgr.hxx (minor upstream cleanup: <cstdio> instead of <stdio.h>, = delete instead of private). - bug1653659.patch: regenerated for v1.7.3 csutil.cxx/.hxx; also adds a MOZILLA_CLIENT stub for unicodeisalpha (returns 0; the function is never reached on the RLBox sandbox path but needs to compile without utf_info.hxx, which is excluded from MOZILLA_CLIENT builds). - bug1739761.patch: refactored, see below. - bug1838113.patch: dropped; upstream incorporated the change in v1.7.3. - bug2037089-build-fixes.patch: new; backports two upstream build fixes needed for Firefox's hazard build. Rebased bug1739761.patch against v1.7.3's significantly restructured HashMgr. v1.7.3 converted several malloc-heavy paths to std::vector and added the H_OPT_OWNFLAGS bit and a release_flags() helper, requiring a non-trivial port. Key invariant: arena_alloc'd memory must only ever see arena_free() (a no-op); calling free()/delete[] on arena interior pointers will crash wasm dlmalloc since they are interior pointers into arena chunks, not the starts of malloc'd blocks. The arena (std::vector<std::unique_ptr<uint8_t[]>>, 64KB chunks) covers hentry structs, flag vectors from decode_flags, flags2 in add_hidden_capitalized_word, and aliasm strings. Runtime additions via add/add_with_affix/remove continue to use new[]/delete[] with H_OPT_OWNFLAGS, which is handled correctly by release_flags. Each arena_alloc call passes its required alignment (alignof(T)), and the allocator pads the bump pointer accordingly. Chunk sizes are rounded up to alignof(std::max_align_t) so that an oversized request (num_bytes >= 64KB) cannot leave a non-aligned current_chunk_size, which would let a later aligned_offset overshoot and underflow the remaining-space check. Differential Revision: https://phabricator.services.mozilla.com/D298926
46 lines
1.6 KiB
HTML
46 lines
1.6 KiB
HTML
<!DOCTYPE html>
|
|
<html>
|
|
<head>
|
|
<title>Test for nsIEditorSpellChecker.suggest</title>
|
|
<script src="/tests/SimpleTest/SimpleTest.js"></script>
|
|
<link rel="stylesheet" href="/tests/SimpleTest/test.css" />
|
|
</head>
|
|
<body>
|
|
<p id="display"></p>
|
|
|
|
<div contenteditable id="en-US" lang="en-US">missspelled</div>
|
|
|
|
<pre id="test">
|
|
<script class="testbody" type="text/javascript">
|
|
add_task(async function() {
|
|
await new Promise(resolve => SimpleTest.waitForFocus(resolve));
|
|
|
|
let { maybeOnSpellCheck } = SpecialPowers.ChromeUtils.importESModule(
|
|
"resource://testing-common/AsyncSpellCheckTestHelper.sys.mjs"
|
|
);
|
|
|
|
let element = document.getElementById("en-US");
|
|
element.focus();
|
|
|
|
await new Promise(resolve => maybeOnSpellCheck(element, resolve));
|
|
|
|
let editingSession = SpecialPowers.wrap(window).docShell.editingSession;
|
|
let editor = editingSession.getEditorForWindow(window);
|
|
let spellchecker = SpecialPowers.Cu.createSpellChecker();
|
|
spellchecker.setFilterType(spellchecker.FILTERTYPE_NORMAL);
|
|
await new Promise(resolve => spellchecker.InitSpellChecker(editor, false, resolve));
|
|
|
|
let suggestions = await spellchecker.suggest("misspelled", 5);
|
|
is(suggestions.length, 0, "\"misspelled\" is correct word");
|
|
|
|
// Hunspell 1.7.3 switched suggestion timeouts from CPU time (clock()) to wall
|
|
// time (std::chrono::steady_clock). On slower builds (sanitizers, 32-bit) the
|
|
// 100ms limit is hit before all 5 suggestions are generated.
|
|
suggestions = await spellchecker.suggest("missspelled", 5);
|
|
ok(suggestions.length >= 3 && suggestions.length <= 5, "\"missspelled\" isn't correct word");
|
|
});
|
|
</script>
|
|
</pre>
|
|
</body>
|
|
</html>
|