Function template inflation in intermediate code. The linker problem we hit with the corrupt file was because there were too many function descriptors and relation in the .lib
The root cause in this bug is that the connection info used by
'SpdyConnectTransaction' is the same instance as the connection info in
'nsHttpTransaction', so we should clone it and let 'SpdyConnectTransaction' use
the cloned one.
The original patch handled the grow case but not the shrink case. When the
current and new allocation sizes are in different size classes, jemalloc's
realloc will move the allocation when shrinking, not just truncate the existing
one.
Based on work by Jon Coppeard.
This involves refactoring the vibrancy and OpenGL/Pixel rendering changes contained
in the following Mozilla meta bugs: 1496823 and 1491445
Also add Big Sur to the features tests and update popup menu look and feel based
on Mozilla bug 1656301.
There don't seem to be any drawbacks to this; tested for the past month disabled
and there have been no issues with any sites visited. Adoption seems very low.
- Explicitly |this| the Asserts in Cell.as<>
- Unified deprot in regexp-error.h
- Convert RegExpShared* inner() to a reinterpret_cast because for SOME reason
.as<> no longer works after the reimport (which is a static cast)
This is required for UniquePtr to accept <void>,
which is required for PseudoHandle = mozilla::UniquePtr<T, JS::FreePolicy>;
in turn for mozilla::SegmentedVector<PseudoHandle<void>> uniquePtrArena_;
Tag #1679
Three debug sources are producing link warnings and will be solved as Issue #1676 progresses.
This does not include JS Testing sources.
Specifically: js/src/gdb or js/src/jsapi-tests
While we do fail a couple of tests, the other mainstream browsers also fail them and I think our implementation of tab-size is good enough to be unprefixed at this point. Having this patch also makes testing easier.
Currently -moz-tab-size only accepts <number> values, and both Chrome and Firefox currently support <length> values and have for some time now. So with this you would be able to support sizes in px or em, for instance. This was implemented in Firefox 53 and was trivial to backport.
This prevents a potential race and simplifies the code a bit by keeping the
bytes read separate instead of using mData, which is modified from another
thread from OnDataAvailable.
Relaxed atomics are fine for these, since they don't guard any memory.
This actually keeps both pseudo-elements for now, since the prefixed version is
still used internally, but we need the unprefixed version for web compat.
Note: while unprefixing a non-spec-compliant pseudo here, it's exactly in line
with what other browsers do. Nobody is following the spec here and at least
we'll be doing what everyone else is with our unprefixed version.
Mozilla's original implementation of this failed a couple of tests, but this seems to solve all the problems. Basically, the caret-color wasn't able to be set differently based on whether a link was visited, and the auto value implementation was incomplete. The only test we fail now is the one where you have grey text on a grey background and the caret is supposed to be visible, but I think that may have been removed from the spec. Even if it wasn't, no other browser supports it anyway.
This CSS property allows input carets (that blinking input cursor you see in text fields), to be given a custom color. This was implemented in Firefox 53, and it was such a minor feature that no one ever missed it, but I don't see any harm in implementing this.
https://bugzilla.mozilla.org/show_bug.cgi?id=1063162
This aligns with the current spec regarding overflow-wrap: break-word and
overflow-wrap: anywhere in if it affects intrinsic sized due to considering
soft-wrap opportunities or not.
See CSS Text Module Level 3, Editor’s Draft, 1 October 2020, Section 5.5
I got very anxious about making sure I included VARIANT_OPACITY in all the places VARIANT_NUMBER was included to make sure it couldn't possibly break unexpectedly, and that led to me accidentally breaking a mechanism that prevented percentages from serializing as numbers in other parts of the code. It was a total accident, and these additions were unnecessary. Basically, the situation is that there was one part of the code where it determines what's allowed for the flex statement (and possibly other statements) by checking whether it got stored as a "number", and basically only disallows percentages if it attempted to store/serialize them as percentages.
However, it only got to that part of the code because I accidentally allowed VARIANT_OPACITY as a valid way for certain tokens to parse where it wasn't necessary. If it tries to parse it that way under very specific circumstances... percentages will be marked valid and fed through the system as numbers rather than being rejected and not serialized at all, because the check to disallow percentages there relied on them being stored as percentages.
It's a really weird thing to have a problem with in a lot of ways, because if percentages aren't allowed in a field, you would think people wouldn't try to use them there, much less depend on the broken behavior that results from them not parsing as a related value.
A race condition seemed to exist between tab destruction and un-registering a ResizeObserver resulting in a null deref crash.
The original reporter in Forum Topic 25311 experienced this on msn.com so that was the functional test reference.
Since these are just interpreted comments, there's 0 impact on actual code.
This removes all lines that match /* vim: set(.*)tw=80: */ with S&R -- there are
a few others scattered around which will be removed manually in a second part.
This make MediaQueryList inherit from EventTarget and adds MediaQueryListEvent
as an interface as well as the onchange() method.
This should not affect compatibility with other code; the event object is a
MediaQueryListEvent instance, which is recognized as a MediaListQuery instance.
This DiD measure ensures that our async HRTF database loading is completed
before we actually try to use it. If not done, database() simply returns null.
Even though percentages are already treated as floats internally by the style system for computation purposes, you have to go out of your way to stop them from being read back out as percentages. What I do here amounts to storing the percentage token in the "wrong" container, the one normally used for floats. This allows a value that was read in as a percentage to be read back out as something else, which is normally prevented by the design of the style system.
This preliminary step allows percentages to be computed and display correctly,
but unfortunately it fails a test after changing VARIANT_HN to VARIANT_HPN because that allows values to be serialized as percentages. However, not doing this means percentages are rejected as valid values for the user to input. The way the style system is setup makes it hard to change this for opacity without changing it for everything else, especially since some code-saving speed hacks in Bug 636029 and Bug 441367 that make a lot of assumptions about this stuff very rigid.
There are situations where nsCSSClipPathinstance->CreateClipPath(dt)
returns null. We need to check for this before trying to use its
functions. If there is no clip path, then always return "no hit".
Host name was purely being used for HPKP and since HPKP is killed,
this can also go. Currently it doesn't do anything other than
generating build warnings.
Since the local variable is always initialized to false, we don't actually need to declare it and can just pass "false" directly as a parameter to the PrepareSheet function's bool. I was worried about code readability at first, but some well-placed comments took care of that.
This clarifies the assumptions the code is making and the order in which the variables pass through the loading process. The new variable is set after the sheet is created and prepared, and is assumed to be false in the beginning.
This is just a clean port of 1322191 and follow-up 1325970. It really seems to add create a new way to access existing code relating to block formatting and floating elements rather than implementing new functionality, and it is mercifully straightforwards.
Unless a user is debugging media errors, this detail is unnecessary to report
and could include sensitive data which could be abused by third-party
requesters. This aligns it with the standard success/error paradigms in normal
browsing situations.
This commit does several things:
- Moves the pref check from ScriptLoader to ns[I]Document so it can be called on
the document.
- Changes the atrribute freezing function to a better name that takes the
document as a parameter.
- Sets the proper async/defer attributes on HTML script elements based on
keywords and whether they are module scripts or not.
This simplifies handling of combinations of async/defer by assigning one and
only one state to scripts.
If async then always async > if defer or module then defer > otherwise blocking.
This makes sure we don't block body-referred sub-resources by head-referenced
defer and async scripts. This is important for all script loads, not just
modules, but is added here because it was run into while implementing modules.
Our use of SQLite explicitly avoids using MSVC's _msize which the lib would
notrmally use. This was set up this way in BZ bug 719579 for legitimate reasons
(because _msize would end up using the system symbol, which would crash trying
to get the size information from jemalloc allocations). Since we build the
SQLite lib separately there is no easy way to avoid this warning as it's not
directly linked against mozglue.dll which has the jemalloc functions it wants
for _msize when not using the system allocator. It is implicitly assuming the
function is an extern returning an int (with the call specifically casting to
int) which is correct (well, correct _enough_, anyway) so the warning can
safely be ignored for our specific use as it all lines up at link time.
Newly introduced aspect-ratio property did not have CSS_PROP_LIST_EXCLUDE_INTERNAL defines, resulting in the following assertion:
\!nsCSSProps::PropHasFlags(p, (1<<28)) (properties defined outside of #ifndef CSS_PROP_LIST_EXCLUDE_INTERNAL sections must not have the CSS_PROPERTY_INTERNAL flag), at ...layout/style/nsCSSProps.cpp:289
This patch resolves the assertion by adding #ifndef around the aspect-ratio property.
This is not very "clean," and is mostly done in the same sloppy way as Emilio did it because that's basically the only way you can do it. Note well that this does NOT actually turn off everything I've done in a clean fashion like ifdefs would. For instance, the Explicitly Enabled flag is still present, but is now always false because the only condition that can set it true is behind the pref and therefore inert when this pref is off. Also, because the arguments of SetDisabled have changed, my modifications to SetMozDisabled must be present regardless of whether the pref is on or off. What I have done is turn off the actual reflection of the disabled attribute in Disabled and SetDisabled, as well as in AfterSetAttr.
However, turning the pref off seems to restore more or less our old behavior, though there may be subtle differences unlike with an ifdef since this is, unfortunately, not an exact science and I can only turn off changes that happen within individual functions and not changes in how functions interact with each other.
This part of the bug was significantly complicated by the following major refactors:
https://bugzilla.mozilla.org/show_bug.cgi?id=1456435https://bugzilla.mozilla.org/show_bug.cgi?id=1459498
As best as I can tell, we just need to implement the explicitly enabled
flag on every instance of GetStyleSheetInfo, make sure
aIsExplicitlyEnabled is false in every situation except the one where
the disabled content attribute is removed from a link element, and
enable alternate stylesheets if this flag is set on them. So we take the
explicitly enabled flag as an input to PrepareSheet, and also add it to
LoadStyleLink and LoadInlineStyle. I also decided not to defer loading of
alternate stylesheets that have been explicitly enabled.
So basically, I'm trying to adapt this to UXP:
https://bugzilla.mozilla.org/show_bug.cgi?id=1281135
The earliest source of difficulty while adapting Bug 1281135 to our codebase was simply getting the new ErrorResult flag added to the SetDisabled function to play nice with the SetMozDisabled function. At this point, the implementation can actually have a stylesheet be disabled by default but there are supposedly issues with alternate stylesheets.
At first I played around with the return type of SetMozDisabled to no avail, but I found another solution fairly quickly.
https://bugzilla.mozilla.org/show_bug.cgi?id=846972https://bugzilla.mozilla.org/show_bug.cgi?id=1157898
Essentially, the way around the problem of the number of return arguments not matching up is to declare a local variable within SetMozDisabled called ErrorResult rv, and using that to store the return value of the ErrorResult argument from SetDisabled. After that, because ErrorCode was removed, you would return rv.StealNSResult() in order to report success or failure to any consumer that calls on SetMozDisabled.
Since this is supported as an alias by Firefox and Edge for the same
reasons and we have websites using this to (attempt to) override the
system-provided styling with their own, leaving out the only supported
keyword we'd otherwise have (with -moz- prefix) but still stating
-webkit-.
TODO: unprefix this completely and make the vendor prefixes aliases.
This is apparently used for fallback selection and if available it is "assumed"
Shadow DOM is also available, while this is a utility function.
Webcompat is a nightmare sometimes.
The added debug assertion does not work due to missing API.
They were added in BZ 1337491, 1395366, and others, but were primarily used for multi-threading.
This uses our existing non-multithreaded syntax instead, resolving a `no member named` build error.
The added debug assertion did not take into account microtask refactoring done in BZ 1405821.
Resulting in error: no member named 'IsInMicroTask' in 'nsContentUtils'.
This resolves the error.
https://bugzilla.mozilla.org/show_bug.cgi?id=1547792
Aspect Ratio handling simplified by using floating point integers:
- Multiplication of value (or inverse value) to a known side for Scaling
- No unequal equal values such as "4/3" vs "8/6" vs "20/15"
- Truly "Empty" aspect ratios, even if one dimension is not 0
At first, HTMLEditor::GetActiveEditingHost might return null in this situation, we should check whether nullptr is returned. At second, SplitNodeDeep returns error since curent is design mode and selection node has no parent. So we should check error.
Ref: Bug 1350772
I meant to do this a long time ago, but basically it accounts for the new XP_SOLARIS build flag that never made it into the MailNews code. Additionally, it enables a compatibility flag for Solaris 11.4 that allows us to use the three-argument implementation of ctime_r still used by Solaris 11.3 and illumos (which also appears equivalent to the NSLDAPI_CTIME implementation used by libldap internally). Also, the ctime_r function has been added to the time.h header library for a while now, not sure why Mozilla thought we didn't have a ctime_r implementation.
When using AF_CONN addresses, don't put these in the COOKIE
chunk. For these addresses it is possible to reconstruct them
locally. Conceptually, addresses are something to be shared
with the peer, but in the case of AF_CONN this might not be
the case. Therefore, zero then out.
Thanks to Natalie Silvanovich of Google Project Zero for
finding and reporting the issue.
JIT optimizations involving the Javascript 'arguments' object could
potentially confuse later optimizations, so we simply disable these
optimizations as a DiD measure.
These should all be spec-compliant and were (for release-trickling of features)
arbitrarily disabled by Mozilla at our fork point. There's no real reason to
keep them disabled since they are used in the wild.
A version bump triggering add-on compatibility check will re-evaluate the "native" status (among other things) so no extra or special work is required to fix erroneous warnings.
tl;dr instafix!
The added debug assertion did not take into account microtask refactoring done in BZ 1405821.
Resulting in error: no member named 'IsInMicroTask' in 'nsContentUtils'.
This resolves the error.
This changes NativeObject::slotSpan() to get the class from the object group
rather than getting it from the base shape to avoid a race between
Shape::ensureOwnBaseShape and background sweeping.
The implementation is based on the work by Bernstein and Yang
(https://eprint.iacr.org/2019/266)
"Fast constant-time gcd computation and modular inversion".
It fixes the old mp_gcd and s_mp_invmod_odd_m functions. The patch also fixes
mpl_significant_bits s_mp_div_2d and s_mp_mul_2d by having less control flow to
reduce side-channel leaks.
Co-authored by : Billy Bob Brumley
Includes removing an error code for a function that never fails, and removing an
error return when the function successfully did what it said it would.
Ref: BZ 1644477
According to the spec this isn't created until the module is instantiated, but
we create it when we compile the module.
We stored this previously in InitialEnvironmentSlot and copied it to
EnvironmentSlot when it was supposed to be created, but we can just store it in
the latter slot straight away and check the module's status and return null if
it shouldn't exist yet.
This reduces the number of slots needed on a moduleObject to 17.
Re: BZ 1420412 Part 1
We can't implement the second part to further reduce our number of slots,
because it relies on SetProxyReservedSlot which in turn relies on
rearchitecturing JS proxies to make reserved slots dynamic. That's a rabbit hole
we really don't want to fall into.
So, we'll end up being a bit slower because it can't be in-line allocated with
having more than 16 slots, but so be it. I sincerely doubt it will make any
practical difference.
This updates module implementation to match spec regarding handling of
instantiation errors, after it was changed yet again, this time to not remember
instantiation errors, but instead immediately rethrow applicable ones.
Ref: BZ 1420420
For checking if a module is in an error state and what the error is, it
shouldn't matter if we are currently GC-ing or not. So we don't need to check
for it, which removes the requirement to pass in the JS context (needed for
AssertHeapIsIdle's runtime check); this unblocks progress where otherwise we'd
have to figure out what the context is at the module level just to satisfy this
check.
This splits ScriptLoader up the same way Mozilla did with the exception of
ScriptRequest due to the fact that ScriptLoader and ScriptRequest are
interdependent and would create a circular dependency if split apart when not
using unified building.
Dynamic script loading/unloading (thank you modules) can yank the script
out from under us before the JS API for it is initialized, leading to
null deref crashes.
This adds a simple check if the passed-in object is sane and present.
Resolves#1602
In a lot of places, edit action handlers and their helper methods return nsresult and aHandled and aCanceled with out params. However, the out params cause the code complicated since:
* it's not unclear if the method will overwrite aHandled and aCanceled value.
* callers need to create temporary variable event if some of them are not necessary.
This patch rewrites the helper methods of HTMLEditRules::WillDeleteSelection() with it.
Tag #1563
When selection is collapsed and JoinBlocks() doesn't handle nor cancel the action, WillDeleteSelection() should move selection to the start/end of leftmost/rightmost editable leaf node and retry to handle the action again.
For avoiding infinite loop, it checks if selected node is changed actually before calling itself again.
Tag #1563
When HTMLEditRules::WillDeleteSelection() tries to remove something from the end/start of a block to its last/first text node but it's contained by block elements, it tries to join the container and the block. However, JoinBlocks() always fails to join them since it's impossible operation. In this case, HTMLEditRules::WillDeleteSelection() should retry to remove something in the leaf, however, it's impossible for now because JoinBlocks() and its helper methods don't return if it handles the action actually.
This patch renames |JoinBlocks()| to |TryToJoinBlocks()| for representing what it is. And this patch adds |bool* aHandled| to the helper methods. Then, *aHandled and *aCancel are now always returns the result of each method. Therefore, for merging the result of multiple helper methods, callers need to receive the result with temporary variables and merge them by themselves.
Note that when they modify DOM node actually or the action should do nothing (for example, selection is across tables), aHandled is set to true.
Tag #1563
Although the headers were removed, they are still referenced in WebIDL causing compiler to fail for debug builds.
This removes TestFunctions.webidl and TestInterfaceJSMaplikeSetlikeIterable.webidl, with nit fix for existing incorrect whitespace.
This removes the (default disabled) node.rootNode readonly attribute
and replaces it with a node.getRootNode() function per WhatWG
spec discussion.
Based on work by John Dai <jdai@mozilla.com>
This is entangled with the IPC messaging configuration so
we need these IPDLs even if sppech synthesis is otherwise
not being built.
This resolves#1560
With the big amount of code churn around DOM a lot of tests
broke severely enough that they caused build bustage.
This commit cleans up, removes or otherwise fixes tests
that are broken, no longer relevant or obsolete.
The JEMALLOC_USES_MAP_ALIGN code has turned out to be worse than useless, and in fact it breaks 64-bit SunOS. It's very old and turned out not to be needed anymore because the way the memory allocator works has changed since it was implemented. It prevented a fix I tried for 48-bit addressing from working properly. However, without either this code or the 48-bit addressing fix, the 64-bit version won't even start, which is why I thought the code was still needed.
https://bugzilla.mozilla.org/show_bug.cgi?id=457189https://hg.mozilla.org/mozilla-central/rev/a26c500b98ab
The 48-bit addressing fix is based on code found here:
https://github.com/OpenIndiana/oi-userland/blob/oi/hipster/components/web/firefox/patches/patch-js_src_gc_Memory.cpp.patch
I already applied these changes to js/src/gc/Memory.cpp, but as I was looking through jemalloc.c I saw that there were very similar ifdefs for Linux on SPARC as the ones I'd had to enaable in Memory.cpp, but for whatever reason the patches I found didn't touch them. So I tried doing for jemalloc.c what was already done for Memory.cpp, and it worked (but only after I removed the map align code).
This removes speech recognition, pocketsphinx, training models
and the speech automated test interface.
This also re-establishes proper use of MOZ_WEBSPEECH to work
for the speech API (synthesis part only) that was a broken mess
before, with some synth parts being always built, some parts
being built only with it enabled and recognition parts being
dependent on it. I'm pretty sure it'd be totally busted if you'd
ever have tried building without MOZ_WEBPEECH before.
Tested that synthesis still works as-intended.
This resolves#1538
Excludes layers/ipc/CompositorBridgeParent and CrossProcessCompositorBridgeParent
because the latter purposely uses functions implemented in the former.
See also code comment in CrossProcessCompositorBridgeParent.cpp
Note: likely dead code for us anyway; likely not worth investing in to solve.
With de-unified building we have to re-apply the fix-up
for <cstdio>, and additionally the same for <cstring>
to prevent the compiler picking the wrong versions of
standard functions.
Before issue #146, border-radius on row groups, rows, column groups, or columns don't apply to the background of each cell, yet the border-radius on the cell itself does.
After issue #146, the behaviors changed. In this patch, I tried to revert the behaviors of border-radius on table row groups, rows, column groups, or columns back to what happened before issue #146.
Also: Don't override GetBorderRadii in nsBCTableCellFrame.
Part 1: Remove current table item, as it's never set.
Part 2: Get rid of generic table painting code, and handle each class separately.
Part 4: Hoist outline skipping into col(group) frame code.
Part 5: Skip box-shadow for table column and column groups.
Part 6: Store column and column group backgrounds separately, and then append them before the rest of the table contents.
Part 7: Pass rects in display list coordinates to AppendBackgroundItemsToTop.
Part 8: Create column and column group background display items as part of the cell's BuildDisplayList.
Part 9: Used cached values instead of calling nsDisplayListBuilder::ToReferenceFrame when possible, since it can be expensive when the requested frame isn't the builder's current frame.
Part 10: Make sure we build display items for table parts where only the normal position is visible, since we may need to create background items for ancestors at that position.
Part 11: Create an AutoBuildingDisplayList when we create background items for table columns and column groups, so that we initialize the invalidation state correctly.
In the printing preview, we create continuous table frame if table is too long to containing in a page. But the default value of NeedToCalcHasBCBorders is false which means we don't calculate HasBCBorders for continuous table frame. Thus, the border collapse is not shown when printing preview.
Bug 1441796 - Part 1: Optimize the (pseudo)-stacking context conditions
Bug 1441796 - Part 3: Reuse the results in nsIFrame::BuildDisplayListForStackingContext() for ChildrenHavePerspective(), IsTransformed(), and Combines3DTransformWithAncestors()
Bug 1512244 - Part 1: Cleanup DescendIntoChild
This reverts m-c Bug 668378 and completely removes Telemetry SQLite IO. As a bonus this fixes a potential crash in newer SQLite versions without the need for updating this useless telemetry shim.
This should ideally not be needed on Mac but compilation fails due to
sizeof `DxgiAdapterDesc` not known.
Didn't face this issue on Linux though, probably due to a different include
chain which already has this file included somewhere along the way.
This de-unifies up to and including dom/media/platforms with the exception of
dom/media/platforms/wmf, because that one throws weird errors in the Win
SDK which don't have a clear cause (yet)
This de-unifies up to and including dom/media/platforms with the exception of
dom/media/platforms/wmf, because that one throws weird errors in the Win
SDK which don't have a clear cause (yet)
This file not only somehow escaped my pass through here but was also mocking me by using AppConstants in a preprocessed file BUT ALSO right AFTER a preprocessor directive.
Reference:
15:03.06 ../../build/unix/gold/ld: error: /home/mattatobin/development/.obj/trunk-master/navigator-x64-gtk3/toolkit/library/../../dom/fetch/Request.o: requires dynamic R_X86_64_PC32 reloc against '_ZN7mozilla3dom9FetchBodyINS0_7RequestEED2Ev' which may overflow at runtime; recompile with -fPIC
15:03.06 ../../build/unix/gold/ld: error: /home/mattatobin/development/.obj/trunk-master/navigator-x64-gtk3/toolkit/library/../../dom/fetch/Response.o: requires dynamic R_X86_64_PC32 reloc against '_ZN7mozilla3dom9FetchBodyINS0_8ResponseEED2Ev' which may overflow at runtime; recompile with -fPIC
15:03.06 ../../build/unix/gold/ld: error: read-only segment has dynamic relocations
15:03.07 /home/mattatobin/development/.obj/trunk-master/navigator-x64-gtk3/toolkit/library/../../dom/fetch/Request.o:Request.cpp:function mozilla::dom::Request::~Request(): error: undefined reference to 'mozilla::dom::FetchBody<mozilla::dom::Request>::~FetchBody()'
15:03.07 /home/mattatobin/development/.obj/trunk-master/navigator-x64-gtk3/toolkit/library/../../dom/fetch/Response.o:Response.cpp:function mozilla::dom::Response::~Response(): error: undefined reference to 'mozilla::dom::FetchBody<mozilla::dom::Response>::~FetchBody()'
Running into the same problem as with dom/canvas: derived template
classes not being happy about being non-unified. Kept unified for now
to prevent linking errors.
KeyFrameEffect and KeyFrameEffectReadOnly are still built unified as
buddies because they have some complex interdependency that couldn't
easily be resolved without linking failures.
* Fixup FindFrameForContentSibling to don't duplicate work and trigger assertions for display: contents
* Enable lazy frame construction for display: contents direct descendants
Tag #1375
* Implementation for assignedNodes
* Include slots in the flat tree
* Fix event get-the-parent algorithm for a node
* Update and add reftests for Shadow DOM v1
* Update web platform tests expectations
Tag #1375
* Cleanup a bit the ShadowRoot code
* Privatize ShadowRoot methods
* When the shadow tree distribution changes, post a restyle + reframe
* Simplify ShadowRoot::IsPooledNode
* Be a bit better at detecting distribution changes
Tag #1375
* Remove eHTMLTags
* De-expose HTML group constants[]
* De-expose gHTMLElements[]
* Split nsHTMLElement
* Clean up nsElementTable.{cpp,h}
* Fixup for eHTMLTag removal in Parser
Tag #1375
This is based on Bug 1395828
* Add nsHTMLElement::IsBlock()
* Rename nsHTMLTags methods
* Remove AssertParserServiceIsCorrect()
* Remove most uses of nsIParserService/nsParserService
* Use inFlowFrame to check the target frame is summary and its parent is details
* Check summary frame instead of summary element on removing the summary
Tag #1375
* Stop calling SetHasDirAuto/ClearHasDirAuto in input element code
* Introduce event state flags that track the state of an element's dir attribute
* Rewrite our existing checks for the state of the dir attr on top of the new event state flags
* Add pseudo-classes for matching on the dir attribute states
* Use the new dir attribute pseudoclasses in html.css
Tag #1375
* Moves side effects of nsGenericHTMLElement and Element's SetAttr, UnsetAttr, and ParseAttribute functions to the corresponding BeforeSetAttr and AfterSetAttr functions
* Moves side effects of HTMLAnchorElement's SetAttr, UnsetAttr, and ParseAttribute functions to the corresponding BeforeSetAttr and AfterSetAttr functions
* Moves side effects of HTMLImageElement's SetAttr function to the corresponding BeforeSetAttr and AfterSetAttr functions
* Moves side effects of SetAttr, UnsetAttr, and ParseAttribute functions to BeforeSetAttr and AfterSetAttr
Tag #1375
* Add a node property to access the ::before and ::after pseudo-elements
* Look for the frame for ::before and ::after pseudos
* Clean up pseudo-element props
* Simplify nsLayoutUtils callers, and make child iterators notice display: contents pseudos
Tag #1375
* Tidy RestyleManager::ContentStateChanged
* Convert UndisplayedMap to a typed hashtable
* Cleanup infallible or unchecked nsCSSFrameConstructor methods
Tag #1375
* Prevent canvas custom content from becoming NAC when reframing the root element
* Add an API to get computed style values through an AnonymousContent object
Tag #1375
* Move the changing of HTMLInputElement's mType from ParseAttribute to AfterSetAttr
* Move a few more things from AfterSetAttr to HandleTypeChange
Tag #1375
* Remove the generic attr preparsing mechanism from BeforeSetAttr and just preparse class attributes directly in the one place that needs to do it
* Move calls to BeforeSetAttr to after AttributeWillChange
* Remove UpdateState calls in BeforeSetAttr
* Move calls to AfterSetAttr to before UpdateState when manipulating attributes
* Remove UpdateState calls from AfterSetAttr, since they are no longer needed there
Tag #1375
* Stop using a node bit for HasExplicitBaseURI
* Move MAY_HAVE_CLASS to mBoolFlags
* Add a flag to indicate that a node is native anonymous content
* Allow tagging of pseudo-implementing native anonymous content with the pseudo type at creation time, and eliminate explicit style contexts in nsIAnonymousContentCreator::ContentInfo
Tag #1375
* EditorBase shouldn't refer mActionListeners directly in loops because it might be removed during a loop
* Create an alias of the type of mEditorObservers
* Create an alias of the type of mDocStateListeners
Tag #1375
* Make StyleChildrenIterator skip NAC generated by root element primary frame ancestors.
* Add nsINode::GetFlattenedTreeParentNodeForStyle.
* Add iterator class to find all restyle roots.
NOTE: Parts 1, 2, and "4.2"
Tag #1375
* Rename nsIDOMEventTarget::PreHandleEvent to nsIDOMEventTarget::GetEventTargetParent
* Add nsIDOMEventTarget::PreHandleEvent
* Add EventTargetChainItem::GetFirstEventTarget
* Call EventTargetChainItem::PreHandleEvent even it sets mCanHandle=false
* Move form control frame focus/blur from nsGenericHTMLFormElement::GetEventTargetParent to PreHandleEvent
* Move fire change event from HTMLTextAreaElement::GetEventTargetParent to PreHandleEvent
* Refine nsXULElement::GetEventTargetParent
* Move dispatch XUL command from nsXULElement::GetEventTargetParent to PreHandleEvent
* Move fire events and set value from HTMLInputElement::GetEventTargetParent to PreHandleEvent
* Add test case
* Let HTMLInputElement delegate event handling to it's parent class
* Refine EventTargetChain flags to reduce overheads
* Refine event target chain creation
* Refine assertion in EventTargetChainItem::Create
Tag #1375
IsSelectionEditable should check whether the focus node and anchor node
aren't null before trying to use them.
This also changes the initialization of selections' aOutIndex to the
last range in the selection as a fallback in case we don't add a range
later (in AddItem) which could also end up with a null selection
otherwise if the additional selection nodes are removed.
To respond dynamically to the pref change, the mAppVersion string needs
to be rebuilt.
Includes some minor improvements for corner cases and removes leftover
b2g junk.
Since we are using vsync for timing anyway in the layout refresh driver,
there is no need to bump the system timer resolution down to 1 ms.
This resolves#1507.
- Adds buildconfig option --enable-nss-sqlstore
- Prefixes NSS dbinit with either sql: or dbm: depending on config
- Pre-initializes mozStorage when NSS-SQL storage is used to prevent
an sqlite3_config race in NSS Init
whether they should be drawn.
This reduces the size of display lists for tables by only adding display
list items that are actually going to be visibly drawn, which will help
overall performance of table drawing.
* So far this seems to only affect Macintosh builds but other unix builds may be affected. Windows doesn't seem to be.
* This maintains consistency with other one-off file invocations and backwards compatibility with older comm setups
There were actually two separate logical errors in this method:
The first part is that "origSizes.isSome()" is simply a bogus
requirement for applying min/max-sizes here. I'm still keeping
the optimization of not needlessly copying the mSizes array
(as originally intended) since it's a quite common case.
The second bug is that min/max-sizes were only applied under
the "if (fr != 0.0f)" block. This is bogus since the calculated
'fr' value depends on 'aAvailableSize' which might change by
applying min/max-sizes and thus 'fr' could become non-zero in
the second round.
To fix, this patch just moves "applyMinMax" block out one level.
Correct various pre-processor defines for sparc64 and in mozjemalloc use the JS arm64 allocator on Linux/sparc64.
This corrects build problems opn Linux sparc64 and is in line with bugzilla bug #1275204.
When setting up connectivity, e.g. with new network detection, the pref
wouldn't be checked (or honored), resulting in spurious startup calls
despite being preffed off. This fixes that.
Resolves#1474
Part 1: Remove current table item, as it's never set.
Part 2: Get rid of generic table painting code, and handle each class separately.
Part 4: Hoist outline skipping into col(group) frame code.
Part 5: Skip box-shadow for table column and column groups.
Part 6: Store column and column group backgrounds separately, and then append them before the rest of the table contents.
Part 7: Pass rects in display list coordinates to AppendBackgroundItemsToTop.
Part 8: Create column and column group background display items as part of the cell's BuildDisplayList.
Part 9: Used cached values instead of calling nsDisplayListBuilder::ToReferenceFrame when possible, since it can be expensive when the requested frame isn't the builder's current frame.
Part 10: Make sure we build display items for table parts where only the normal position is visible, since we may need to create background items for ancestors at that position.
Part 11: Create an AutoBuildingDisplayList when we create background items for table columns and column groups, so that we initialize the invalidation state correctly.
This should be the last code backout for this. merging this branch
should get us back to the way we were (+ additional code changes for
later changes) as fasr as the unused unboxed code is concerned.
This adds an addition to the environment set up for child processes
(plugin container) so that it may still be able to pass the omni
parameters there as-needed.
This stub is added because websites insist on considering this
very hardware-dependent and O.S.-variable low-level font-control
as a "critical feature" which it isn't as there is 0 guarantee
that font variation settings are supported or honored by any
operating system used by the client.
On top this is a WD status feature that sites shouldn't be using, and
the feature itself is strongly discouraged for use in favor of standard
CSS font manipulation keywords like `font-weight`.
This is a hack to deal UnifiedBindings trying to deal with an incomplete codegen implementation so it was worked around with a hack back in Firefox 30-something. If we have anymore of this non-sense crop up as stuff is added or removed from DOM then extend this conditional.
Least until something better comes along.
Do less work in PaintRowGroupBackgroundByColIdx for cells that are not in our desired set of columns.
crash fix:
Guard against empty column groups when building a display list for a table.
If EME isn't built, these functions would have no use and the callback
structures are unavailable. Stubbing these out will still allow the
callers to work unchanged, but in effect simply always return without
dispatching a promised task.
This fixes the dependency issue from part 2a.
HTML standardizes proper behavior of colSpan and rowSpan:
The main thing is that getting the .rowSpan and .colSpan IDL properties will now return the actual clamped value that we use.
There is flexibility in exactly the value the initialized length must
hold, i.e. if an array is completely empty, it is valid for the
initialized length to be any value between zero and the length of the
array, as long as the in-memory values below the initialized length have
been initialized with a hole value.
In the case of 0, the array is packed and the move operation would be a
nop, so simply convert the assert to a condition to save some cycles.
This was added as part of the WebExtension sec issue CVE-2017-5389.
Since we no longer have this interface, it needed to be removed to
prevent errors in the WebRequest channel callback.
Since we are dealing with the element (nodeInfo->LocalName() and NameAtom() are the same value), we could use nodeInfo->NameAtom() instead.
Tag UXP Issue #1344
This is all the manual work for Bug 483155, minus the added functionality to disable SVG and MathML which can be done at any time and are out of scope.
Tag UXP Issue #1344
Per spec, document objects have a throw-on-dynamic-markup-insertion counter, which is used in conjunction with the create an element for the token algorithm to prevent custom element constructors from being able to use document.open(), document.close(), and document.write() when they are invoked by the parser.
Tag UXP Issue #1344
This is a follow-up patch for bug 1392970. Since we only set CustomElementDefinition on a custom element which is custom, we could add more assertion to ensure that.
Tag UXP Issue #1344
We make nsContentUtils::EnqueueLifecycleCallback static so that it can be called without a window object. To achive this, we also make CustomElementReaction not taking a CustomElementRegistry in the constructor, as it can call Upgrade statically.
Tag UXP Issue #1344
We call attributeChangedCallback in two cases:
1. When any of the attributes in the observed attribute list has changed, appended, removed, or replaced.
2. When upgrading an element, for each attribute in element's attribute list that is in the observed attribute list.
Note: w/ Fixup for not implementing an API Enhancement Bug 1363481.
Tag UXP Issue #1344
There are two places doing prototype setup in old upgrade,
- If definition comes after JS reflector creation, CustomElementRegistry::Upgrade will do prototype swizzling.
- If definition comes before JS reflector creation, Element::WrapObject will set up the prototype.
The later one does SubsumesConsideringDomain, but the former doesn't not.
This patch is to fix the inconsistency, i.e. the former case should also do SubsumesConsideringDomain.
Tag UXP Issue #1344
1. It is possible that invoking a reaction triggers pushing a new ElementQueue into ReactionStack (e.g., calling define() in constructor which probably enqueue another upgrade reaction), and the reference of ElementQueue passed to InvokeReactions becomes invalid due to the memmove in nsTArray implementation.
2. And we get another benefit from this is memmove becomes faster.
Tag UXP Issue #1344
Bug 1347446 makes accessing ElementReactionQueue becomes a bit non-trival (have to get it via DocGroup). Since bug 1359346 already refactors the creation time of CustomElementData, ReactionQueue can also be put into CustomElementData, then we can just get ReactionQueue from Element.
Tag UXP Issue #1344
There are two changes here:
1) We allow setting .body even if the root element is not an <html:html>. This is what the spec says to do, and what we used to do before the changes in bug
366200. No tests for this yet, pending https://github.com/whatwg/html/issues/3403 getting resolved.
2) We use GetBody(), not GetBodyElement(), to look for an existing thing to replace. This matters if there are <frameset>s involved.
Tag UXP Issue #1344
Tag UXP Issue #252
This gets rid of platform-dependent hard-coded defaults, but keeps
build-time blocking if there is no GL provider (in which case layers
acceleration almost certainly won't work because it needs a GL
compositor and would likely crash without)
New prefs are
- layers.acceleration.enabled to enable HWA
- layers.acceleration.force to force it enabled (requires .enabled to be
set as well)
This is the platform part of this issue. The rest will be front-end work
(Preference UI integration and pref migration)
LIST and RETR still appear to work as intended on ftp:// URLs after my changes. I wasn't able to test STOR because the browser doesn't appear to support FTP uploads at this time (although our FTP implementation appears perfectly capable of doing an FTP upload.)
If I understood the issue correctly, though, what we're doing is ensuring that we receive a preliminary 100 response from the FTP server for a given action before jumping to the 200 code describing what we do if the action was completed. Even though it makes no logical sense for a server to say an action was completed before it was initiated, someone could write a really annoying FTP server that takes advantage of this fact to crash the browser if they wanted.
Our NSS version is closer to the currently-released .1, so bump version
to that.
Note: we still have some additional patches to the in-tree version in
place so this isn't a 100% match to the RTM one.
The indicated BZ bug was resolved in Gecko 50, and could have already
been enabled before (since it returns a promise as-required).
With the rest of promise-based media implemented it makes no sense to
keep this disabled on production.
Considering the one exception this was still enabled for on all versions
of Firefox since 45 until our fork point is no longer applicable.
This resolves#1353
We've kept the java source up-to-date until its removal, so there should
be very little additional java mangling needed to have it back up to
speed and usable again.
This reverts commit c6446f1126.
module is loaded afterwards,
Summary: When the builtin trust module is loaded after some temp certs
being created, these temp certs are usually not accompanied by trust
information. This causes a problem in UXP as it loads the module from a
separate thread while accessing the network cache which populates temp
certs.
This change makes it properly roll up the trust information, if a temp
cert doesn't have trust information.
- Store the clipboard even if it was set in a GTK dialog.
- Fix a GtkTargetList leak.
- Notify GTK that the data is no longer available for
clipboard_get_cb(), so that GTK will no longer advertise nor attempt to
store the data.
Microsoft IME on Windows 10 20H1 (build 19025+) supports IME private
mode by input scope. Although previous Windows versions use an
undocumented API for Edge and IE only, the next Windows 10 release will
use a public API for it.
We pre-empt this potential privacy concern by setting the IS_PRIVATE
flag when in PB mode.
This NSS version (and onward) has variables to enable PKDBF rounds for
master password hardening (combating brute-force guessing).
Since we use DBM, we need to explicitly enable iteration counts in DBM
mode, and adjust the rounds used with a lower maximum because the PKDBF
routine for DBM is considerably slower for which the NSS default of
10,000 is debilitatingly high.
The method for this is run-time configuration through environment
variables. We set these prior to NSS initialization from nsAppRunner.cpp
so the NSS lib picks them up and allows us to continue using the DBM
cert/key stores with stored rounds per record.
This prevents mail applications from attempting to use an unauthenticated connection to mailbox(s) and avoid unexpected deletion of local mbox files and subsequent re-download of mailbox content over imap.
https://bugzilla.mozilla.org/show_bug.cgi?id=1244768
I happened to find an older version of the promise-based media playback patch in Bugzilla, the one that was originally submitted for review. It had the DocShell changes I already knew how to deal with, and had fewer of the audio wrapper and nsISupports changes that were confusing me in the later patch. I was able to do a better job getting this back into a UXP-appropriate configuration than I could have with the final version.
I'm honestly still a little unsure about some of the minor refactoring done in the patch itself, insisting on already_AddRefed promises and such, but I don't really know how to avoid those completely. Still, I think it's better than it was.
CheckForOutdatedParent()
This was only used to check for cases when document.open() changed the
global, and elements being inserted into the document needing a new
reflector as a result.
Since document.open() no longer changes the global, this code is no
longer needed.
This changes the work we do for document.open() in the following ways:
- We no longer create a new Window when doing document.open().
We use the same Window but remove all the event listeners on the
existing DOM tree and Window before removing the document's existing
children to provide a clean slate document to use for .write().
- We no longer create a session history entry (previously would be a
wyciwyg URI). We now replace the current one, effectively losing the
entry for the original document.
- We now support document.open() on windowless documents.
This updates our behavior for computed DOM styling to no longer return
null on elements that have no display, but return a 0-length (empty)
style instead and don't throw. For this we stop looking at having a
presentation for the style and just look at the document instead.
This resolves#1219
This add a few disabled compiler warnings to the compiler flags on MSVC
and specifically filters out the RTTI disabling CL flag to prevent CLI
override warnings.
This resolves#1316
Previously-used `Courier New` was purposefully used at a smaller font
size due to its width. With the font switch this forced downscaling is
no longer needed, although it should still remain somewhat smaller than
variable width for balance (variable width is inherently more compact).
- Converts from macro to static const for the limit.
- Uses a check against the declared type for the counter instead of
a hard-coded one.
This resolves#1317.
SFTKSession objects are only ever actually destroyed at PK11 session
closure, as the session is always the final holder -- and asserting
refCount == 1 shows that to be true. Because of that, NSC_CloseSession
can just call `sftk_DestroySession` directly and leave
`sftk_FreeSession` as a no-op to be removed in the future.
forming a monotonically increasing sequence (e.g. with position:sticky)
Relative positioning can cause table parts to move, which can cause
issues with the cursor position to know which rows can be skipped.
To make this work, use the max difference between the frame's rect and
the union of the frame's "normal" position and the overflow rect to
cover the area of relatively positioned elements even if they are out of
order.
This resolves#1309.
list collisions when processing the background color of a table.
This is effectively #146 Part 3c to go along with the other 2 there.
This resolves#1308.
"USING VLA'S IS ACTIVELY STUPID! It generates much more code, and much
slower code (and more fragile code), than just using a fixed key size
would have done." -- Linus Torvalds
We never use jemalloc without MOZ_MEMORY surrounding code (mostly a
FreeBSD concession which doesn't work anyway). This allows us to get rid
of the "poor man's mutex" (CPU-based spinlock) in mozjemalloc.
This basic validation is hard-set to always-on, so no point in keeping
it conditional since we basically always want this rudimentary pointer
validation to be done.
- Check for undefined/null regex flags (because...)
- Make it throw a typeerror instead of syntax error on non-global
- Generalize JS error messages for these checks.
* The langpack is not generated via the package target but IS for mozpackage target using old behavior
* Add mach command to specifically generate the langpack when needed without the rest of the packaging routine (mach langpack)
* This adds what will eventually be a rewritten packaging routine while maintaining backwards compatibility with the original mozilla routine
* Changes the build target and installer makefile to use the alternative packaging
* Adds build target to specifically invoke the original mozilla routine
* Update mach commands accordingly
I didn't include this in my initial PR because I assumed this part
was not relevant to UXP. I was asked to submit this while the MailNews rebase
was going on, but I was late with my patch and didn't quite grasp that this
specifically was still needed.
My counterpart at Mozilla, Petr Sumbara, has gotten his version of this
approved.
https://bugzilla.mozilla.org/show_bug.cgi?id=1568473
I made a typo in commit 687a789 when updating the ifdef style I used to comply with UXP standards. The typo I made resulted in a compiler warning I failed to notice, so I was asked to tag issue #457 when submitting the PR. I also fixed some trailing whitespace I apparently left behind in the file.
This provides a charity-supporting search engine option focused on
the planet's health and animal welfare.
Since this engine is based on Bing search, it provides a balancing
alternative to Ecosia (Google-based) with similar eco goals.
This function never writes beyond `aDest` + `aMaxOutputSize`, and is
therefore protexted against malicious datapackets.
It also ignores unconsumed input upon reaching `aMaxOutputSize` and can
therefore be used for parial decompression of LZ4 input up to a desired
resulting size of decompressed data.
MOZ_ASSERT changed to NS_WARNING. Also correctly handle case where last chunk ends with \r. Tested to make sure that regression identified in Bug 1494764 comment 10 remains fixed and that non-chunked and chunked messages are handled correctly including when \r\n is split between chunks.
Tag #1273
* Fix undefined HRESULT codes in class CMapiApi in MapiApi.cpp.
* Fix doubled up sub-expressions in MIME and MAPI.
* Remove broken/unused morkRowCellCursor::MakeCell()
Tag #1273
* strip unnecessary ==== padding in base64 text.
* Extension of the characters to be encoded (quoted-printable) according to RFC2047.
* Remove unwanted characters from headers. Compact extraneous white space in display name to avoid sender address spoofing.
* Be more tolerant of spaces in base64-encoded RFC 2047 tokens.
Tag #1273
Change mainly affects Windows. However, Linux integration also changed so that it also allows new mail notification to occur for folder with SentMail flag.
Tag #1273
Suspected "use after free" in nsMsgLineStreamBuffer::ReadNextLine() leading to crash since object may be destroyed while still in use on another thread.
Tag #1273
MOZ_ASSERT changed to NS_WARNING. Also correctly handle case where last chunk ends with \r. Tested to make sure that regression identified in Bug 1494764 comment 10 remains fixed and that non-chunked and chunked messages are handled correctly including when \r\n is split between chunks.
Tag #1273
When cross-compiling for Windows on Linux, fix mismatched types in nsMsgImapSearch.cpp. nsString::get() returns a char16ptr_t (with MOZ_USE_CHAR16_WRAPPER) and that causes a mismatch so we cast it to a const char16_t*. usAsciiCharSet is declared as a nsAutoString so we instead declare it as a char16_t* to avoid subsequent type mismatches.
Tag #1273
Also factors out some common code into nsMSgDBFolder::Delete(), and makes a couple of incidental changes to try and clarify the responsibilies of nsIMsgPluggableStore::DeleteFolder().
Tag #1273
* Part 1: make conversion CTORs explicit.
Minor tweaks to conform with new(ish) static analysis rule which flags up implicit single-argument conversion constructors.
* Part 2: fix a couple of double-mRefCnt declaration errors.
* Part 3: Fix unsafe use of NS_ConvertASCIItoUTF16().
Tag #1273
This only applies to non-literal+ appends. Bad response typically occurs when {size} parameter of append command is larger than server can accept. This prevents TB from sending megabytes of data to a server that has already rejected it.
Tag #1273
This is because it was only used in response to a CAPABILITY imap request. However, often CAPABILITY response occurs without a request so TB sees no need to do a request. Fix by moving the usage of use_literal_plus to where LITERAL+ capability is actually used.
Tag #1273
- Use HEAD instead of GET for probe to avoid loading pages
- Reduce retries to 2
- Reduce timeout to 10 s (since we're just getting a HEAD this is royal)
- Identify ourselves to websites as an automated tool
- Improve performance of list merging (O(n^2) was getting too expensive)
- Add a total counter and perform GC every 200 requests
- Move header licensing from tri-license to MPL 2.0. MPL-compatible
other licensing has been retained where originally present.
- Remove individual superseded licensing terms.
- Remove patches, outdated readmes & incomplete patch summaries.
- Remove incomplete cairo release notes (only went up to 1.6.4 anyway).
- Rewrite COPYING to indicate the current state of the library in tree.
This also adds a reftest for border radius on collapsed borders (should
be ignored according to the CSS3 standard). We didn't do this before,
except on internal elements.
list items.
This patch does the following things:
1. Creates nsDisplayTableBorderCollapse that draws all collapse border
of tables.
2. Stops the use of nsDisplayTableBorderBackground.
3. Lets column and column group frames generate display items.
4. When traversing the table, also traverses the column and column group
frames.
5. For each type of table frame (col group, col, row group, row and
cell), draws their own background.
This removes a ton of tests that are no longer relevant with (un)watch
removed (e.g. testing stability/bugs in the watchpoint system itself
which has never been the most stable), and updates others that would
previously rely on watch/unwatch, so that they don't unexpectedly fail.
HKDF-Expand enforces a maximum output length much shorter than stated in
the RFC. This patch aligns the implementation with the RFC by allocating
more output space when necessary.
On MacOS, particularly large allocations within the platform limits
(1.2G+) will fail and crash. This adds a specific size check for that
when working around driver bugs (default).
While there, added a generic size_t limited size check for the
platform, and reporting OOM if too large.
HKDF-Expand enforces a maximum output length much shorter than stated in
the RFC. This patch aligns the implementation with the RFC by allocating
more output space when necessary.
On MacOS, particularly large allocations within the platform limits
(1.2G+) will fail and crash. This adds a specific size check for that
when working around driver bugs (default).
While there, added a generic size_t limited size check for the
platform, and reporting OOM if too large.
This also addresses clearing of document dependent JS slots which might
get out of sync with innerWindow navigation; relevant comments added.
This resolves#1253
This also addresses clearing of document dependent JS slots which might
get out of sync with innerWindow navigation; relevant comments added.
This resolves#1253
This should do it for all the commits to files I changed, but while I'm in here I could probably go ahead and turn ALL the singular if defined statements into ifdef statements by using grep/find on the tree. On the other hand, perhaps we should do that as a separate issue so that this doesn't become a case of scope creep.
OpenIndiana didn't need this for some reason, but on Oracle Solaris, we need this to make sure we're using gsed (GNU sed) here. It's probably a safer bet anyway.
This fix is a bit ugly and may need to be changed later if we switch a new GCC version, but the fact is that we use an architecture-specific path for GCC libraries on Solaris, so knowing the right prefix for GCC would only help so much, because it would still need to decide between ${gccdir}/lib and ${gccdir}/lib/amd64. The MOZ_FIX_LINK_PATHS variable puts the search paths into the right order without the need for me to use elfedit on the binaries afterwards.
https://developer.mozilla.org/en-US/docs/Mozilla/Developer_guide/Build_Instructions/Compiling_32-bit_Firefox_on_a_Linux_64-bit_OS
Setting this up turned out to be easier than I thought it would be. All I had to do was apply these instructions in reverse and add the following to my .mozconfig file:
CC="gcc -m64"
CXX="g++ -m64"
AS="gas --64"
ac_add_options --target=x86_64-pc-solaris2.11
export PKG_CONFIG_PATH=/usr/lib/amd64/pkgconfig
ac_add_options --libdir=/usr/lib/amd64
ac_add_options --x-libraries=/usr/lib/amd64
Most of these changes were fairly trivial, just requiring me to make a few of the changes I made earlier conditional on a 32-bit build. The biggest challenge was figuring out why the JavaScript engine triggered a segfault everytime it tried to allocate memory. But this patch fixes it:
https://github.com/OpenIndiana/oi-userland/blob/oi/hipster/components/web/firefox/patches/patch-js_src_gc_Memory.cpp.patch
Turns out that Solaris on AMD64 handles memory management in a fairly unusual way with a segmented memory model, but it's not that different from what we see on other 64-bit processors. In fact, I saw a SPARC crash for a similar reason, and noticed that it looked just like mine except the numbers in the first segment were reversed. Having played around with hex editors before, I had a feeling I might be dealing with a little-endian version of a big-endian problem, but I didn't expect that knowledge to actually yield an easy solution.
https://bugzilla.mozilla.org/show_bug.cgi?id=577056https://www.oracle.com/technetwork/server-storage/solaris10/solaris-memory-135224.html
As far as I can tell, this was the last barrier to an AMD64 Solaris build of Pale Moon.
https://bugzilla.mozilla.org/show_bug.cgi?id=1397928
Was looking into that _POSIX_PATH_MAX/NAME_MAX issue earlier because it didn't make a lot of sense and I was thinking of other approaches besides char arrays, and I wanted to make sure it didn't cause problems after they did it. Turns out that one commit after this was added, Mozilla determined the code I was working on fixing to be dead code as of Firefox 58. I don't know if it's dead code in Pale Moon as well, but given that it compiles fine without it and I can't find any other references to szExeFile in the IPC code, that seems like a safe bet.
Besides, I determined config/pathsub.c already seems to do what this code looks like it's trying to do, and implements the solution of just defining NAME_MAX to 256 and having done with it that I nearly adopted after realizing that even OS/2 and BeOS, let alone Unix/Linux systems, all basically use that value and there's just disagreement on which system header to check for it.
https://github.com/oracle/solaris-userland/blob/82dd4adb0eca729372074d62435e00a783d95b1f/components/desktop/firefox/patches/firefox-49-npapi.patch
The first fix was something I found on Oracle's patchset and allowed me to use the last Flash Player compiled for Solaris, from all the way back in 2012. Still works with most Flash content.
The second is an evolution of what I had to do to get Interlink to compile. For Interlink, I basically had to copy the contents of any boolean values from confvars.sh into the empty moz.configure file, otherwise nothing would get configured. I decided to test whether Pale Moon had the same issue, and it turned out that it wasn't as bad as on Interlink, but it was still pure luck that the browser component built at all, because MOZ_PHOENIX and other important flags were apparently not being defined at all, hence why I couldn't get half the extensions to be compatible at first.
I don't know why this is the case, but apparently configure.in isn't able to import values from confvars.sh. old-configure.in seems immune to the problem that application-specific configure.in files were experiencing, though. confvars.sh itself seems to work fine with values that aren't passed along via configure.in, though. So it's the interface between those two files that is messed up.
https://www.tachytelic.net/2019/01/grep-recursively/
During testing, I tried simply replacing grep with ggrep, which was non-portable but worked fine. Using an environment variable with os.getenv set to 'g' also worked, but the problem with that approach is that you have to set it manually and some times the script will mess up if you don't explictly define it to an empty string for platforms that don't need it.
Setting TOOLCHAIN_PREFIX to 'g' seemed to solve all of my problems except this one, and it appears to be the only non-portable use of GNU grep in the whole tree. To understand what I tried to do here, let me present with you with two commands that yield the same output on my machine:
find . -name '*.xul' | xargs grep -E "#include"
ggrep -E -r "#include" --include="*.xul" .
I just tried to make the Python script follow the logic of those two commands, though I'm not sure how well I succeeded since I got no errors. I visualized everything this way:
ggrep -E -r "<strings>" --include="<filesuffixes>" <path>
find <path> -name '<filesuffixes>' | xargs grep -E "<strings>"
And arranged it all accordingly to the best of my ability, though I should point out that Python is not my strong suit.
https://bugzilla.mozilla.org/show_bug.cgi?id=1185424http://www.mindfruit.co.uk/2012/06/relocations-relocations.html
The libxul.so fix was implemented by Mozilla in Firefox 57 and personally recommended to me by an Oracle employee on the OpenIndiana mailing list. It can easily be made ifdef XP_SOLARIS, but it seems like the new way is considered a better solution overall by the original author of the code that had it use that null pointer hack to begin with.
I can't link where I found the fix for libffi because I came up with it myself while looking at the way sysv.S does things. Something clicked in my brain while reading that mindfruit link above, though, and gave me enough of a sense of what was going on to be able to fix libffi.
The libffi fix looks a bit hairy because of all the FDE_ENCODE statements, but if you examine the code closely, you should find that it does exactly what it did before on all platforms besides Solaris. I later discovered that people who originally ported Firefox to Solaris never figured this out during the Firefox 52 era and had to use GNU LD for linking libxul.so while using the Sun LD for the rest of the build to make it work. For whatever reason, it works for me now without the GNU LD trick.
https://bugzilla.mozilla.org/show_bug.cgi?id=1513913
Mozilla's solution to this is arguably overkill, since the namespace issue on Solaris only required them to change (or temporarily undefine) __log2. Instead they changed ALL the functions to be something along the lines of dbm_log2. They haven't changed the external interface at all, though.
If you're unhappy with this patch, I think I could also use XP_SOLARIS ifdefs to undefine __log2 prior to where it's declared in the dbm headers. The good thing about Mozilla's solution is that it guarantees this namespace issue never occurs again on any platform, though.
https://bugzilla.mozilla.org/show_bug.cgi?id=1350355
u_int32_t is not an stdint.h type. Windows already requires this, Solaris needs it too. If someone has a nit with this approach, the alternatives include:
1. Just replacing every instance of u_int32_t with uint32_t.
2. Including <sys/types.h> for Solaris only, which does define this.
3. Changing the original ifdef to be WIN32 || XP_SOLARIS
But it really doesn't matter how you solve this problem, all of the approaches are functionally equivalent, and this one has been used in Firefox since version 55. As far as I can tell, all it does is apply a fix that was being done for Windows already to any platform that needs it.
https://bugzilla.mozilla.org/show_bug.cgi?id=1375467
I would ifdef this, but it's been in Firefox since version of 56, and Petr Sumbara's explanation as to why it's wrong in the first place is so detailed that it's pretty obvious the code wasn't technically doing things properly to begin with.
Basically, they tried to redefine a system function after including the header file that declares it, and it caused problems on Solaris because libc functions are imported into the C++ std namespace in a different way that also complies with standards. So the existing implementation is technically bad code on all platforms, the Solaris implementation just uncovered the lack of standards compliance in the Mozilla code.
https://bugzilla.mozilla.org/show_bug.cgi?id=1345102
I assess this change to be low-risk for the following reasons:
1. It has been in Firefox since version 55 without issues.
2. The current behavior is not POSIX compliant, and is retained in the one instance where the new functionality causes issues.
3. It makes safer assumptions about implementation details than what we have now.
https://bugzilla.mozilla.org/show_bug.cgi?id=1351309https://bugzilla.mozilla.org/show_bug.cgi?id=1309157
I assess this change to be low-risk for the following reasons:
1. It has been in Firefox since version 55 without issues.
2. It's nearly identical to the fix for bug 1309157 which is already in our tree, so that one would be causing problems if this one were going to.
3. Linux, Windows, and BSD are known to have a hypot function in their math libraries.
4. Even if it does break something, it should only break a test and not critical functionality.
https://bugzilla.mozilla.org/show_bug.cgi?id=1352449
Mozilla patch that's been in the code since Firefox 55. Seems like there have been no ill effects from implementing it, and it would only increase the portability of the UXP code. All the Solaris Firefox repos I've seen implement some variation on the jsexn patch, and this seems to be the cleanest version of it.
I can add ifdefs if needed or there are performance concerns associated with this patch, but I get the impression this alignment backlog issue might affect a few platforms other than Solaris, though none were named. Otherwise I think they wouldn't have used "platforms that need it" in plural form or failed to ifdef it.
Current versions of libcubeb already have a Sun audio implementation, but in Firefox 52 and earlier, this was all they had. I'm not completely happy with this implementation because it has issues like video freezing if the soundcard isn't working, but I think fixing this or pulling in a newer libcubeb would be going too far for too little gain.
https://bugzilla.mozilla.org/show_bug.cgi?id=1364865
Solaris doesn't define NAME_MAX because if you read the current POSIX standard literally, no system that supports multiple file systems or networking should be defining it. It's a pedantic choice given that they USED to define NAME_MAX, but Solaris always did take POSIX compliance seriously, for better or worse.
This is mostly ifdefs, but as you can see, Solaris is actually a lot like Linux. They're both more SysV than BSD at core, and most of the differences have more to do with Solaris not using glibc than anything else.
I still need to audit a lot of these changes and understand why they're needed and what the alternative approaches are. After this patch, most of the core functionality needed to build Solaris is here.
https://bugzilla.mozilla.org/show_bug.cgi?id=535106https://bugzilla.mozilla.org/show_bug.cgi?id=1359841
Like many parts of the busted Solaris support, this one has its origins in the pre-Firefox days. Bug 535106, another Mozilla suite bug. It keeps coming up because the core issue is never addressed, the fact that nsTSubstring doesn't know how to handle pid_t. I think the explicit cast to int is a band-aid they use because they know if they touch that substring header file to make it handle pid_t, they'll probably be asked to fix all the other problems with it. I honestly think it just works by accident on other platforms because it's implicitly cast to signed or unsigned int, even though the POSIX standard says pid_t can be either long or int, and work as either a signed or unsigned integer. Whatever the case may be, it's handled better on Solaris now than it was.
Ironically enough, the main point of having pid_t rather than just having pids be int or something is to hide this little implementation detail so you can just use pid_t for the return type in portable code without having to worry about what it is on a specific platform. The unfortunate way Mozilla implemented string functions turns that on its head and makes the good things about pid_t into liabilities rather than assets.
https://bugzilla.mozilla.org/show_bug.cgi?id=1360571
Solaris uses ksh as the default shell, and furthermore bash doesn't seem to support the 'local' keyword feature when invoked as sh on OpenIndiana. We could just change the script to use bash (it is an option even on Solaris), but this fix is available and Mozilla has been using it since Firefox 55 with no issues on any other platforms. It was specfically done this way to avoid any need to change existing mozconfig files, so I feel confident saying the change is totally benign and if anything the way it is now is technically a POSIX compliance issue inherited from Mozilla that we'll hit if we ever try to compile this on any Unix platform where bash isn't sh.
Compared with what Pale Moon had for Solaris originally, this is mostly the same zero point I started patching from, but I've made the following changes here after reviewing all this initial code I never looked at closely before.
1. In package-manifest.in for both Basilisk and Pale Moon, I've made the SPARC code for libfreebl not interefere with the x86 code, use the proper build flags, and also updated it to allow a SPARC64 build which is more likely to be used than the 32-bit SPARC code we had there.
2. See Mozilla bug #832272 and the old rules.mk patch from around Firefox 30 in oracle/solaris-userland. I believe they screwed up NSINSTALL on Solaris when they were trying to streamline the NSS buildsystem, because they started having unexplained issues with it around that time after Firefox 22 that they never properly resolved until Mozilla began building NSS with gyp files. I'm actually not even sure how relevant the thing they broke actually is to Solaris at this point, bug 665509 is so old it predates Firefox itself and goes back to the Mozilla suite days. I believe $(INSTALL) -t was wrong, and they meant $(NSINSTALL) -t because that makes more sense and is closer to what was there originally. It's what they have for WINNT, and it's possible a fix more like that could serve for Solaris as well. Alternatively, we could get rid of all these half-broken Makefiles and start building NSS with gyp files like Mozilla did.
3. I've completely cut out support for the Sun compiler and taken into account the reality that everyone builds Firefox (and therefore its forks) with GCC now on Solaris. This alone helped clean up a lot of the uglier parts of the code.
4. I've updated all remaining SOLARIS build flags to the newer XP_SOLARIS, because the SOLARIS flag is no longer set when building Solaris.
5. I've confirmed the workaround in gtxFontconfigFonts.cpp is no longer necessary. The Solaris people got impatient about implementing a half-baked patch for a fontconfig feature that wasn't ready yet back in 2009, and somehow convinced Mozilla to patch their software to work around it when really they should have just fixed or removed their broken fontconfig patch. The feature they wanted has since been implemented properly, and no version of Solaris still uses the broken patch that required this fix. If anyone had ever properly audited this code, it would have been removed a long time ago.
When Mozilla implemented this initially, only Windows 8 existed.
Because of the strict equal check, 8.1 and 10 didn't get HWA over RDP
while they are perfectly capable of doing so with RemoteFX. This change
allows any version of Windows from 8.0 onwards to use HWA over RDP.
360 Safeguard/360 Total Security (Qihoo) causes crashes in a11y components.
This adds the offending dll to the injection blocklist.
See BZ bug 1536227 for details.
Add NSS_CMSSignedData_GetDigestAlgs and NSS_CMSSignedData_hasDigests which are required for security patches in mailnews applications. Ref: m-c bug 1526473
Add NSS_CMSSignedData_GetDigestAlgs and NSS_CMSSignedData_hasDigests which are required for security patches in mailnews applications. Ref: m-c bug 1526473
When Mozilla implemented this initially, only Windows 8 existed.
Because of the strict equal check, 8.1 and 10 didn't get HWA over RDP
while they are perfectly capable of doing so with RemoteFX. This change
allows any version of Windows from 8.0 onwards to use HWA over RDP.
List of relevant patches applied:
1398537 part 2 - [css-multicol] Implement percentages for 'column-gap' (Gecko part).
1434478 part 1 - [css-grid] Stop back-computing percentage grid gaps when the percentage basis is indefinite. Treat them as zero sized instead.
1434478 part 2 - Stop back-computing percentage padding/margin when the percentage basis is indefinite. Treat them as zero sized instead.
1434478 part 3 - Remove IntrinsicISizeOffsetData::hPctPadding/hPctMargin members since they are now unused.
1434478 part 4 - Factor out constants like NS_UNCONSTRAINEDSIZE so they can be used in headers without needing nsIFrame.h (idempotent patch).
1434478 part 5 - Create nsLayoutUtils::ResolveToLength for resolving CSS <length-percentage> (idempotent patch).
1434478 part 6 - Propagate a percentage basis to nsIFrame::IntrinsicISizeOffsets for resolving padding/margin.
This is needed only for CSS Grid since in other cases we're only using IntrinsicISizeOffsets in the inline-axis and the percentage basis is always indefinite for *intrinsic sizing*. When calculating the intrinsic size of grid items in the grid container's block axis however, we do have a definite size for the grid area in the inline-axis and it should be used per: https://drafts.csswg.org/css-grid/#algo-overview "2. Next, the track sizing algorithm resolves the sizes of the grid rows, using the grid column sizes calculated in the previous step." (Percentage padding/margin for grid items is always resolved against the grid area's inline-size nowadays.)
List of relevant patches applied:
1425599 part 15 - [css-grid] Test reference fixes + more tests.
1373678 Part 3: Add line number checks to test_grid_implicit.html.
1416350 - Part 3: Add test to verify line numbers of grids with leading implicit tracks.
1416350 - Part 4: Add a reftest of repeat:auto-fit grids with leading implicit tracks.
1417711 - [css-grid] An abs.pos. grid container child that only covers removed 'auto-fit' tracks should not span to the end padding edge.
1416350 - Part 5: Correct the expected results for grids that have leading implicit tracks.
1418727 part 3 - [css-grid] Reftest updates.
List of relevant patches applied:
1425599 part 1 - [css-grid] Change the track sizing algorithm for spanning items so that it accumulates individual item contributions to the plan by max() rather than incrementing the planned size directly.
Also, fix a bug when copying back the planned limits after updating it for the first span group. It should only copy back track sizes that were actaully spanned by those items, other content-sized tracks' limits should remain at "infinity".
1425599 part 2 - [css-grid] Factor out the min-sizing parts of the track sizing for spanned items to a templated method (idempotent change).
1425599 part 3 - [css-grid] Factor out most of the max-sizing parts of the track sizing for spanned items to a templated method (idempotent change).
1425599 part 4 - [css-grid] Factor out the starting base/limit size to a templated method (idempotent change).
1425599 part 5 - [css-grid] Make CollectGrowable a templated method so that it works with either base/limit sizes (idempotent change).
1425599 part 6 - [css-grid] Make the size distribution methods templated with the intent of merging them in a later patch (idempotent change).
This patch also introduces an eInfinitelyGrowable bit to help get rid of the 'limits' temporary track sizes in the next patch.
1425599 part 7 - [css-grid] Remove the 'limits' copy of track sizes since they are no longer needed (idempotent change).
1425599 part 8 - [css-grid] Factor out the fit-content clamping function from DistributeToTrackLimits and pass it as a param instead (idempotent change).
1425599 part 9 - [css-grid] Merge DistributeToTrackLimits/Bases (idempotent change).
1425599 part 10 - [css-grid] Make MarkExcludedTracks a static method since it doesn't use 'this' (idempotent change).
1425599 part 11 - [css-grid] Hoist the marking of excluded tracks out from GrowSelectedTracksUnlimited to a separate method (idempotent change).
1425599 part 12 - [css-grid] Merge CopyPlanToBase/Limits into a templated method instead (idempotent change).
1425599 part 13 - [css-grid] Merge Grow[Base|Limits]ForSpanningItems into a templated method instead (idempotent change).
1425599 part 14 - [css-grid] Use iterators instead of an array + start/end index for the item data (idempotent change).
1425599 part 16 - [css-grid] Make SizeContributionForPhase a template.
1425599 - [css-grid] Follow-up bustage fix for stupid compiler warnings.
1378481 - Assign 'roundingError' in the default branch too, to avoid a maybe-uninitialized compiler warning.
1423292 - [css-grid] Add a couple of ItemState bits to Dump(), and make an ItemState assertion stricter (DEBUG-only changes).
1373678 Part 1: Reduce grid line numbers by count of leading implicit lines, minimum 0.
1416350 - Part 1: Correctly account for removed 'auto-fit' tracks also when there are leading implicit tracks.
1416350 - Part 2: Correct logic for Grid API line numbering with leading implicit tracks.
1418727 part 1 - [css-grid] Introduce StateBitsForRange() that collects the union of the state bits for a range of tracks (idempotent change).
1418727 part 2 - [css-grid] Require that an item spans at least one track with an 'auto' min sizing function for Automatic Minimum Size to apply.
360 Safeguard/360 Total Security (Qihoo) causes crashes in a11y components.
This adds the offending dll to the injection blocklist.
See BZ bug 1536227 for details.
Although it's enabled by default in gcc (should be, anyway!), we're
being explicit here for 64-bit x86 platforms here also. This matches the
old behavior.
This avoids getting data synchronously on the main thread in an XHR
(which has been deprecated for a long time and _may_ actually be blocked
in our networking) and attempts to be more predictable by always firing
an update request for the dialog from the XHR request handlers.
null, if we're doing a replace.
We're going to end up hitting this if someone does a document.open()
before mOSHE has been set. We shouldn't need to worry about mLSHE,
because the document.open() will cancel the corresponding load.
This patch implements:
- noModule getter/setter for HTMLScriptElement
- the nomodule attribute for HTMLScriptElement
- the logic in nsScriptLoader that denies the loading of a nomodule script
- tests
Tag #618
In addition to further development based on the Mozilla upstream code, and
selective cherry-picking of directly-applicable patches, this repository has its
own development and holds the base for a future platform to be used by XUL
own development and holds the base for a maintained platform to be used by XUL
applications.
This repository will contain at least one application to demonstrate and make use
of the platform: The Basilisk web browser, a close twin to Mozilla's Firefox.
For a list of active projects making use of the Unified XUL Platform, checkout http://thereisonlyxul.org/.
## Additional documentation
@@ -19,7 +18,8 @@ directory. This will contain relevant documentation regarding contributing,
using and distributing this code and its binaries.
If you are interested in the development and building side of things, some
information will be available on the [Pale Moon developer wiki](http://developer.palemoon.org).
information will be available on the [Pale Moon developer site](http://developer.palemoon.org).
You are also always welcome to get in touch with our community on the [Pale Moon forum](https://forum.palemoon.org/).
### A note about trademarks and branding
@@ -29,7 +29,11 @@ owners. For more details, please see the notifications in the respective directo
### Foundation and maintainership
This repository has been founded and is maintained by Moonchild (M.C. Straver).
If you fork this repository to perform your own work on it, please consider
offering improvement patches upstream to its origin to mutually improve the
platform and build a future for XUL.
This repository has been founded by Moonchild (M.C. Straver) and is maintained by him
and other community members.
If you fork this repository to perform your own work on it, please consider offering
improvement patches upstream to its origin to mutually improve the platform and build
a future for XUL.
## Looking for Pale Moon and Basilisk?
As of [`Issue #969`](https://repo.palemoon.org/MoonchildProductions/UXP/issues/969), release versions `28.9.0` and `2020.03.04` respectively, they have been split off into their own dedicated repositories located at [`MoonchildProductions/Pale-Moon`](https://repo.palemoon.org/MoonchildProductions/Pale-Moon) and [`MoonchildProductions/Basilisk`](https://repo.palemoon.org/MoonchildProductions/Basilisk).
# This Source Code Form is subject to the terms of the Mozilla Public
# License, v. 2.0. If a copy of the MPL was not distributed with this
# file, You can obtain one at http://mozilla.org/MPL/2.0/.
Loaded 100 of 21061 files, more files were not shown because too many files have changed in this diff.
Show more
Reference in new issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Moonchild
Matt A. Tobin
Werner Lemberg
Henri Sivonen
David Major
Alessio Vanni
Brian Smith
Matt A. Tobin
adesh
athenian200
Gaming4JC
adeshkp
J.C. Jones
Kris Maglione
Jon Coppeard
Andy
Michael Tuexen
Andrea Marchesini
Jan de Mooij
Sohaib ul Hassan
Byron Campen [:bwc]
wolfbeast
Lootyhoof
win7-7
Fedor
Ascrod
Trushita
Cameron McCormack
Jiaxun Yang
Bas Schouten
Moonchild
JustOff
athenian200
JMadgwick
Simon Giesecke
sotaro
Kai Engert
Daiki Ueno
Yaron Tausky
Makoto Kato
Chris Peterson
Craig Disselkoen
Christian Holler
Jonathan Kew
Valentin Gosu
David Parks
Mats Palmgren
André Bargull
Kevin Jacobs
Dmitry Grigoryev
wolfbeast
janekptacijarabaci