summaryrefslogtreecommitdiff
path: root/xfa/fxfa
AgeCommit message (Collapse)Author
2018-08-01Replace more unchecked casts with type conversion funtions.Tom Sepez
Change-Id: If3eb4396232daed28f2b63ffe1d110075c9299de Reviewed-on: https://pdfium-review.googlesource.com/39230 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
2018-08-01Rename PageSetContainerLayoutItem to TraverseStrategy_PageSet.chromium/3509Tom Sepez
Provides consistency with the naming of other traverse strategies. Fix dubious casting inside the strategy. Change-Id: Ic96b1922ae1dc601a37c8c3b3dd0f118ab6b2eea Reviewed-on: https://pdfium-review.googlesource.com/39210 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
2018-08-01Add a ScopedLocale for use within CXFA_LocaleValue.Lei Zhang
Add more checks for |m_pLocaleMgr| before dereferencing it. Change-Id: I39ab44d652364f5530266d2b724fa6703d1b51f1 Reviewed-on: https://pdfium-review.googlesource.com/39114 Commit-Queue: Lei Zhang <thestig@chromium.org> Reviewed-by: Ryan Harrison <rharrison@chromium.org>
2018-08-01Fix a nullptr crash in CXFA_LocaleValue::ValidateValue().Lei Zhang
Fix some nits as well. BUG=chromium:868271 Change-Id: Ia3231fde98c3e16e41b092a9833402cedc8e828d Reviewed-on: https://pdfium-review.googlesource.com/39112 Reviewed-by: Ryan Harrison <rharrison@chromium.org> Commit-Queue: Lei Zhang <thestig@chromium.org>
2018-07-31Remove CXFA_Corner downcasts from cxfa_rectangle.cppTom Sepez
We don't need anything that the parent class can't provide, so avoid dubious casting. Change-Id: I9f3d9735f3bfbcf005865772af31817c2b8addef Reviewed-on: https://pdfium-review.googlesource.com/39270 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
2018-07-30Replace m_bIsContentLayoutItem with proper enum for type.Tom Sepez
Because bools are bogus. Change-Id: I54a2627443784cec40912288230e657c63a35408 Reviewed-on: https://pdfium-review.googlesource.com/39190 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
2018-07-30Don't static_cast<>(this) to subclass in CXFA_LayoutItem methods.chromium/3508Tom Sepez
Instead, move the methods to the subclass where they belong. Fix IWYU in CJX_Object.h to fix compilation. Change-Id: I4c71f28235b9cf5000e9ddaf33d6602baf22205f Reviewed-on: https://pdfium-review.googlesource.com/39170 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
2018-07-30Return unique_ptr from CXFA_FFNotify::OnCreate{Container,Content}LayoutItem()Tom Sepez
... and then immediately release it, but it is a step in the right direction. Change-Id: Ib52972c6789d8f98a576d1c69f8019541c96ac51 Reviewed-on: https://pdfium-review.googlesource.com/39152 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
2018-07-30Convert some asserts to ifs in cxfa_ffnotify.cpp.Tom Sepez
We'll get a nice safe segv should this somehow happen in the wild instead of a type confusion error. We'll assert later on in the debug builds when the null is seen, as well. Change-Id: Iabd8468adcbacaa0acdc7c68f27cc8f94e0e68cc Reviewed-on: https://pdfium-review.googlesource.com/39151 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
2018-07-30Type-safe iteration over CXFA_ContentLayoutItems.Tom Sepez
There's no reason that every CXFA_LayoutItem must be a CXFA_ContentLayout item, so use checked conversions. The code may already be arranging things so the constraint is already satisfied, but this adds an additional layer of safety. Tidy some long expressions with pre-existing typedef for node iterator while we're at it. Change-Id: If1448131ca73bfedb2266b070f1a913688598e64 Reviewed-on: https://pdfium-review.googlesource.com/39051 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
2018-07-30Fix a misspelling of "locale".Lei Zhang
Change-Id: I3fdc6d473f08835b6caeacbbd583bb651a9c5678 Reviewed-on: https://pdfium-review.googlesource.com/39111 Commit-Queue: Ryan Harrison <rharrison@chromium.org> Reviewed-by: Ryan Harrison <rharrison@chromium.org>
2018-07-30Remove a NOTREACHED() in FormValueNode_SetChildContent().Lei Zhang
It is reachable, and the action is currently a no-op, so let it be. Also change FormValueNode_SetChildContent() to return void since no caller checks the return value. Remove its default argument, since all the callers set it. Change-Id: I742ca2d936872a3b3d609c178321199abc06546c Reviewed-on: https://pdfium-review.googlesource.com/39115 Commit-Queue: Ryan Harrison <rharrison@chromium.org> Reviewed-by: Ryan Harrison <rharrison@chromium.org>
2018-07-30Remove CFGAS_PDFFontMgr::m_FDE2PDFFont font cache.Henrique Nakashima
This cache is never read. Change-Id: Iaa4dcf8debf01ab783c2e757a19b18c0317d6fe1 Reviewed-on: https://pdfium-review.googlesource.com/39150 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Henrique Nakashima <hnakashima@chromium.org>
2018-07-28Remove CXFA_TraverseStrategy_ContentLayoutItem.chromium/3507chromium/3506Tom Sepez
It makes illegal casts and requires its callers to double-check the type validity of results before use. Instead, use the parent class iterator and perform checked casts. No functional difference, since it looks like the requisite checks were being made in all places. Make one "using" visible to other files to save some verbosity. Change-Id: I894ca15b4bdddd4723b787663950a58bc58b7f06 Reviewed-on: https://pdfium-review.googlesource.com/39030 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Lei Zhang <thestig@chromium.org>
2018-07-28Add ToXML{Instruction,Text,CharData}() checked conversion functionsTom Sepez
All usages were previously checked correctly, but this consolidates some code as well. Change-Id: I63711748b31b698a3f21f98fdb536db1e9e0b1cf Reviewed-on: https://pdfium-review.googlesource.com/39010 Commit-Queue: Lei Zhang <thestig@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
2018-07-26Use moar ToXMLElement() in place of static_cast<>.Tom Sepez
Introduces checks in a few new places, but mainly just consolidates checking/casting logic. Change-Id: I634a03060d254db099972c6978249992367e146c Reviewed-on: https://pdfium-review.googlesource.com/38900 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
2018-07-25Remove CFX_MemoryStream uses in tests.Lei Zhang
Replace with CFX_BufferSeekableReadStream, which allows for spans and const inputs. Change CXFA_DocumentParser to take IFX_SeekableReadStream instead of IFX_SeekableStream in the process. Change-Id: I0168451350c9fc250231f0414c38738a4d86ca42 Reviewed-on: https://pdfium-review.googlesource.com/38852 Commit-Queue: Lei Zhang <thestig@chromium.org> Reviewed-by: Ryan Harrison <rharrison@chromium.org>
2018-07-25Disable submit in XFA forms.Henrique Nakashima
Bug: chromium:842503 Change-Id: If411815d8324929f482e3cad0fda54f24d370c2a Reviewed-on: https://pdfium-review.googlesource.com/37830 Reviewed-by: Tom Sepez <tsepez@chromium.org> Commit-Queue: Henrique Nakashima <hnakashima@chromium.org>
2018-07-25Introduce ToXMLElement() checked downcast helper functionTom Sepez
And use it in place of ASSERT() that the type was correct. Since we can check it, doing so at runtime may help avoid type confusion, and we'll get a nice safe SEGV if the asserted condition ever fails. Change-Id: I46b65a4b70e819a670d7cad7966e0d100f0d9a63 Reviewed-on: https://pdfium-review.googlesource.com/38810 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
2018-07-25Move CXFA_ThisProxy helper to CXFA_Object.Tom Sepez
Because the other helpers are declared here. Rename VariablesThis to ThisProxy in a few places; VariablesThis is a slightly different concept (see GetVariablesThis()). Then introduce helper for CXFA_List subclass as well. Remove unused const version of some helpers. Change-Id: Ia328d8cd170a8b97015e98c1c770fa8a44810455 Reviewed-on: https://pdfium-review.googlesource.com/38670 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
2018-07-24Fix UAF in CPDFSDK_Widget::GetMixXFAWidget().chromium/3502Henrique Nakashima
Do not allow instanceManager methods to run in Foreground XFA forms. They are static, and their widgets should not be inserted or removed. See "XML Forms Architecture (XFA) Specification Version 3.3", page 272. Bug: chromium:860697 Change-Id: Ia96834e085ee508618ca4dcb2bd5271466369ede Reviewed-on: https://pdfium-review.googlesource.com/38751 Reviewed-by: Tom Sepez <tsepez@chromium.org> Commit-Queue: Henrique Nakashima <hnakashima@chromium.org>
2018-07-23Tighten up ThisProxy casts.Tom Sepez
Previous CLs have shown that the "lpClass" checks aren't sufficient here, so ensure we are always checking C++ enum value before downcasting this type. Change-Id: I418127c5e7131e0a3363363a60d1976719d6837c Reviewed-on: https://pdfium-review.googlesource.com/38550 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
2018-07-23Remove unused CScript_DataWindow::JSDataWindow().Tom Sepez
Do the same thing for N other similar files. Change-Id: If9bb7f6335287441206083c3512fc8a788e9d39b Reviewed-on: https://pdfium-review.googlesource.com/38532 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
2018-07-23Replace CFXJSE_HostObject::type_ with polymorphic conversion methodsTom Sepez
More foolproof technique, and saves one word in every instance at a cost of two pointers in every subclass vtable. Change-Id: Ied4f91d4d203b5e6d83abe751159eb7e55489bb5 Reviewed-on: https://pdfium-review.googlesource.com/38530 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
2018-07-20Replace downcasts in cxfa_node to checked conversion methods.Tom Sepez
Part of the effort to not trust downcasts in the CFXA_Node hierarchy. The number of subclasses is small enough here that the O(n^2) space overhead for the vtables is not significant. Put all sub-classes in same top-level namespace as the parent, just because we avoid jumping in and out of anonymous namespace. Change-Id: I32b80a93c75f8a09eb17c01fe1e29d04b4053fa1 Reviewed-on: https://pdfium-review.googlesource.com/38510 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
2018-07-19Use pdfium::as_writable_bytes() in cxfa_xmllocale_unittest.cppTom Sepez
This is a cleaner way of getting the required argument. Add helper function and anonymous namespace as misc. cleanups along the way. Change-Id: I600074b7f6597ffd9b451a21ee374f937a8e9940 Reviewed-on: https://pdfium-review.googlesource.com/38310 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
2018-07-17Limit recursion depth for CXFA_DocumentParser::NormalLoader().Henrique Nakashima
Bug: chromium:849143 Change-Id: I973bb3be6151ac3afad850533cb735c03e9f3d2c Reviewed-on: https://pdfium-review.googlesource.com/38210 Reviewed-by: Ryan Harrison <rharrison@chromium.org> Commit-Queue: Henrique Nakashima <hnakashima@chromium.org>
2018-07-16Use UnownedPtr/Optional in cfxa_layoutcontext.cppTom Sepez
Introduce cfxa_layoutcontext.cpp to satisfy chromium style ctor check. Fix typo in class forward declaration in cfxa_layoutcontext.h Change-Id: I2b29bf61a21be0f895faf8984897c5655a4f0e89 Reviewed-on: https://pdfium-review.googlesource.com/37950 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
2018-07-16Fix crash when typing letters into an XFA datetime field.Henrique Nakashima
The root of the issue is that CXFA_FFDateTimeEdit inherits from CXFA_FFTextEdit and methods in the former treat its widget as a CFWL_Edit, while it can be a CFWL_DateTimePicker. Bug: chromium:857521 Change-Id: I764b6c03095b16f6a9cf72ff36768ca4c57c4070 Reviewed-on: https://pdfium-review.googlesource.com/37910 Reviewed-by: Ryan Harrison <rharrison@chromium.org> Commit-Queue: Henrique Nakashima <hnakashima@chromium.org>
2018-07-16Process data changes regardless if they can be formattedRyan Harrison
UI elements in XFA can have two different <picture> clauses with different meanings. One is a formatting statement that dictates the format the underlying data should appear in. The other <picture> occurs within an <validate> block and specificies what format the underlying should be. If it doesn't fit this <picture> the <validate> block indicates what should occur. In the existing code actually running the second <picture> type was being conditionally guarded on whether of not the formatting of the first type succeeded. This CL moves the calls related to data changing earlier, to before the formatting and UI updating occurs. BUG=pdfium:1065 Change-Id: Ib525c490074187c3272f4263ab451bf347b74f9c Reviewed-on: https://pdfium-review.googlesource.com/37250 Reviewed-by: Henrique Nakashima <hnakashima@chromium.org> Commit-Queue: Ryan Harrison <rharrison@chromium.org>
2018-07-16Use UnownedPtr in CXFA_LocaleMgrTom Sepez
Change-Id: I74b1828570e54a3868a77231a8ac13bc412c6c43 Reviewed-on: https://pdfium-review.googlesource.com/37051 Reviewed-by: dsinclair <dsinclair@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
2018-07-05Clean up ProcessFormatTestValidateRyan Harrison
The existing implementation was overly complex and hard to understand, so this simplifies the logic. As part of this a number of issues, such as the lack of a failure path, have been resolved. The spec implies that a picture clause of "" should only accept the empty string, but existing implementation returns that validation does exist. This is due to the GetPicture() method returning a string, instead of a pointer or an Optional, so there is no mechanism to differentiate between a clause with an empty string and the clause not being present. This CL maintains the existing behaviour, because there may be code elsewhere that depends on it. The existing implementation returns validation not existing if the node under test is non-interactive. Though this seems intuitively correct, it is problematic, because this logic isn't just called via an interaction with the node. The validity check could be initiated by a call from JS attached to an event or a different node. Thus it is possible for a node under test to be non-interactive, but the result of the test is still important. The caller may make the node interactive if it isn't valid, for example. The spec doesn't state anything about interactivity controlling if validity checks run. This CL removes the related logic, since it was causing issues with an example PDF from the wild. BUG=pdfium:1065 Change-Id: Icb5b97e0d90f6fbc7ad6b87d81e256803c757eb0 Reviewed-on: https://pdfium-review.googlesource.com/37131 Commit-Queue: Ryan Harrison <rharrison@chromium.org> Reviewed-by: Henrique Nakashima <hnakashima@chromium.org>
2018-07-03Use unowned ptr to Node from outside XFA node tree.chromium/3481Tom Sepez
Change-Id: Idcd992224fc5411c97f3001293af83a6c581e695 Reviewed-on: https://pdfium-review.googlesource.com/36991 Reviewed-by: dsinclair <dsinclair@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
2018-07-03Use UnownedPtr<> in cxfa_nodeiteratortemplate.hTom Sepez
Change-Id: I48c20a56ca25a6821bc9b51aa59c037fe2341194 Reviewed-on: https://pdfium-review.googlesource.com/36972 Reviewed-by: dsinclair <dsinclair@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
2018-07-03Use more UnowendPtr<> in cfgas_formatstring.h and cfx_break.hTom Sepez
Move initializers to .h file (proves none missing). Change-Id: Iff8fe76e46634aa32a32ecd592476594a9d1bafa Reviewed-on: https://pdfium-review.googlesource.com/36950 Reviewed-by: dsinclair <dsinclair@chromium.org> Commit-Queue: dsinclair <dsinclair@chromium.org>
2018-07-03Use UnownedPtr<> in cxfa_{ffwidgethandler,imagrenderer,loadercontext}.hTom Sepez
Change-Id: I25496c9792a99ca27d2ab84c267770ff941cf921 Reviewed-on: https://pdfium-review.googlesource.com/36912 Reviewed-by: dsinclair <dsinclair@chromium.org> Commit-Queue: dsinclair <dsinclair@chromium.org>
2018-07-03Use more UnownedPtr<> in cfxa_ffpageviewTom Sepez
Change-Id: Ifdb9fb23d61082b7c1b9b93c057e0881959cea2a Reviewed-on: https://pdfium-review.googlesource.com/36530 Commit-Queue: dsinclair <dsinclair@chromium.org> Reviewed-by: dsinclair <dsinclair@chromium.org>
2018-07-03Remove XFA_LAYOUT_INVALIDNODETom Sepez
XFA_LAYOUT_INVALIDNODE is used in place of nullptr, but does not appear to represent a sentinel value distinct from it. Using a non-nullptr value prevents using any kind of smart pointer type in the future. Change-Id: Iab6418b3dc9c3d45780730795496ac07b67de696 Reviewed-on: https://pdfium-review.googlesource.com/36651 Reviewed-by: dsinclair <dsinclair@chromium.org> Commit-Queue: dsinclair <dsinclair@chromium.org>
2018-07-02Comment some raw pointers that cant become UnownedPtr<>.Tom Sepez
These appear in compile-time const data. Adding the comment makes it easier to see what work remains when using grep. Change-Id: Ibb5fe920dc9fc99dd4c866f9ddf800a58f6a67a8 Reviewed-on: https://pdfium-review.googlesource.com/36690 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
2018-06-29Use UnownedPtr<CXFA_Node> or comment raw pointers as tree nodes.chromium/3479chromium/3478Tom Sepez
Part 2. Change-Id: I26fe434dac3c3f5186414440acc2a495a5f14091 Reviewed-on: https://pdfium-review.googlesource.com/36670 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
2018-06-29Use UnownedPtr to CXFA_Node from outside the treechromium/3477Tom Sepez
Comment raw pointers subject to nondeterministic tree destruction order as such to avoid re-attempting to convert to the unowned mechanism. Change-Id: Ia9fe3c8a2729dc1e2b1de4a8c62ae3d2c3d7ec0a Reviewed-on: https://pdfium-review.googlesource.com/36635 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
2018-06-29Use more UnownedPtr<> in cxfa_ffwidgetTom Sepez
Change-Id: Iae962711f522ad52e77f19201d00d9ab86096ca4 Reviewed-on: https://pdfium-review.googlesource.com/36510 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
2018-06-27Remove CXFA_FField::UnloadWidget().Tom Sepez
It is only called during the destructor, and does things that the destructor would automatically do, like calling into the superclass. So just use destructors. Change-Id: I1196817798eeca99914b44ddee1f75781e682211 Reviewed-on: https://pdfium-review.googlesource.com/36231 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: dsinclair <dsinclair@chromium.org>
2018-06-27Use unowned ptr for IFWL_ delegatesTom Sepez
Change-Id: I1e9059242dad254394b73c6fddb010b182e969f0 Reviewed-on: https://pdfium-review.googlesource.com/36230 Commit-Queue: dsinclair <dsinclair@chromium.org> Reviewed-by: dsinclair <dsinclair@chromium.org>
2018-06-22Avoid some duplicate const/non-const getters.Tom Sepez
Getters should nearly always be const when they return a non-const pointer to an unowned object. Saves a bit of code along the way. Grep for similar occurrences and add "const". Change-Id: I492bf962a7d62452fa40310146226dc0c8ebb753 Reviewed-on: https://pdfium-review.googlesource.com/35890 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
2018-06-21Clean up constant values for JS alert and beepRyan Harrison
Define constant values in the public API for the valid values of alert button type, alert icon type, and beep type. Replace various magic numbers through out the code base using these values. Also replace the XFA specific versions with an enum class that is guaranteed to have the same values, instead of #defines that just happen to. This CL does not attempt to add error checking on these values, since it currently doesn't exist so adding it may cause regressions. Change-Id: Ief3aee2a4ad419691c18fc1dba8b984ad222141b Reviewed-on: https://pdfium-review.googlesource.com/35730 Commit-Queue: Ryan Harrison <rharrison@chromium.org> Reviewed-by: dsinclair <dsinclair@chromium.org>
2018-06-21Remove WideString::c_str() calls where possible.Tom Sepez
Many of these cause a string duplication as we go from Widestring => c_str => Widestring Change-Id: I0dfa952e66f89138e719ff9200db3d9397839e28 Reviewed-on: https://pdfium-review.googlesource.com/35790 Reviewed-by: dsinclair <dsinclair@chromium.org> Commit-Queue: dsinclair <dsinclair@chromium.org>
2018-06-19Move fxcrt::{Byte,Wide}Strings with std::move().chromium/3466Tom Sepez
Remove some string copies in barcode that were noticed whilst looking for moves. Change-Id: Ieda34d00f633576ba1f0dca283dcdabfb36f236c Reviewed-on: https://pdfium-review.googlesource.com/35410 Reviewed-by: dsinclair <dsinclair@chromium.org> Reviewed-by: Tom Sepez <tsepez@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
2018-06-14[formcalc] Calculate length of string when calling FXSYS_wcstofDan Sinclair
When calling the FXSYS_wctof method we currently pass in -1 from AdvanceForNumber. This tells the method to calculate the string length. This can be slow for a formcalc string with a lot of numbers. This CL changes the call to pass in the length of remaining data in the original string. This takes the MSAN runtime of the case in the linked bug from ~21seconds to ~500ms. The debug runtime goes from ~2s to ~500ms. Bug: chromium:846104 Change-Id: Idbd19a728160f35982e21c0d97567fbbeefe667a Reviewed-on: https://pdfium-review.googlesource.com/35210 Reviewed-by: Tom Sepez <tsepez@chromium.org> Commit-Queue: dsinclair <dsinclair@chromium.org>
2018-06-14Move definition of XFA_SCRIPTATTRIBUTEINFO into cjx_object.h.Peter Collingbourne
This allows compilers targeting the MS ABI to select the correct inheritance model for the member function pointer type XFA_ATTRIBUTE_CALLBACK using the complete type of CJX_Object. It will allow us to enable the new Clang flag -fcomplete-member-pointers globally. Bug: chromium:847724 Change-Id: I90cedde8c5355e5eb896a93f0e43e6a1e1d09dbc Reviewed-on: https://pdfium-review.googlesource.com/35190 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Tom Sepez <tsepez@chromium.org>