summaryrefslogtreecommitdiff
path: root/core
AgeCommit message (Collapse)Author
2017-04-18Libtiff upstream: _TIFFcalloc additionNicolas Pena
Upstream commit: https://github.com/vadz/libtiff/commit/d60332057b9575ada4f264489582b13e30137be1 Bug: chromium:711638 Change-Id: I46de1a00f9bb8d5de8df64ec78a9d62dcb4352ed Reviewed-on: https://pdfium-review.googlesource.com/4310 Reviewed-by: Tom Sepez <tsepez@chromium.org> Commit-Queue: Nicolás Peña <npm@chromium.org>
2017-04-18Clean up unneeded code in agg and skia driverNicolas Pena
- The agg driver never initializes an ICC transform and always uses an alpha_flag equal to 0 in a couple of methods. - The skia driver is not using CFX_Renderer. If it needs it later, we should move it out of fx_agg_driver.cpp into its own file, instead of duplicating the code. Change-Id: Ibd721cf1760270bea3f8f95d6e1e8edbc6ba1792 Reviewed-on: https://pdfium-review.googlesource.com/3952 Reviewed-by: Tom Sepez <tsepez@chromium.org> Reviewed-by: Cary Clark <caryclark@google.com> Commit-Queue: Nicolás Peña <npm@chromium.org>
2017-04-18Check char sign in fx_ext.hNicolas Pena
The std methods assume that the input can be represented as unsigned char. Bug: pdfium:703 Change-Id: I11d20869502aff0ebb8badca853398a7d0232338 Reviewed-on: https://pdfium-review.googlesource.com/4171 Commit-Queue: Nicolás Peña <npm@chromium.org> Reviewed-by: Tom Sepez <tsepez@chromium.org>
2017-04-18Cleanup XML buffer codeDan Sinclair
This CL removes the CXML_DataStmAcc class which was unused and moves the CXML_DataBufAcc to a separate file. This removes the use of the IFX_BufferedReadStream as the parent class and removes default implemented methods. Change-Id: I5a3921d5a6f882d54cc1c74305f1a777aa4cebfc Reviewed-on: https://pdfium-review.googlesource.com/4292 Reviewed-by: Tom Sepez <tsepez@chromium.org> Commit-Queue: dsinclair <dsinclair@chromium.org>
2017-04-18Use Byte/WideString iteratorsTom Sepez
Change-Id: I85c8423c177fd7ecd5da90ef89419efc0f9cf44b Reviewed-on: https://pdfium-review.googlesource.com/4262 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
2017-04-18Remove unused classDan Sinclair
The CFindFileDataW class is never used. Change-Id: I230e1e1f2a5cc6a498ac104d156b9f685324a707 Reviewed-on: https://pdfium-review.googlesource.com/4293 Reviewed-by: Tom Sepez <tsepez@chromium.org> Commit-Queue: dsinclair <dsinclair@chromium.org>
2017-04-18Render image masks individually instead of rendering the entire page asrbpotter
a bitmap in order to reduce spool sizes. BUG=682156, 674771 Change-Id: Ic743ba729a7b9609be18fb620a5eb6bc440aed6a Reviewed-on: https://pdfium-review.googlesource.com/2916 Commit-Queue: Lei Zhang <thestig@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
2017-04-17Add iterators for Byte/WideStringCs.Tom Sepez
Change-Id: I40ec07c0da54bcf36c83fa26ff457cd4b98a91cf Reviewed-on: https://pdfium-review.googlesource.com/4261 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
2017-04-17Fix buffer management issues in CFX_SAXReader.Tom Sepez
Re-write to use std::vectors rather than self-managed buffers. Includes test case that breaks before patch. Formerly, we had two independent buffers whose position were tracked by the same variable, assuming that only one was being written to at a given time. This is a bad idea because it's easy to forget to zero the index when switching buffers, and start indexing into one using previously computed offsets from the other. Additionally, there were cases where the location of a partial entity wasn't discarded when switching states thus clearing the buffer tracking said entity. Bug: 711459 Change-Id: I008f69517d4319a5fe8abda8d54c5b9975551697 Reviewed-on: https://pdfium-review.googlesource.com/4230 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
2017-04-17Replace indirect calls with switch in CFX_SaxReader.Tom Sepez
Split off from a forthcoming CL to make it smaller. Change-Id: I4409e8c7e87c0954a909e7c1dbeace9d4435b472 Reviewed-on: https://pdfium-review.googlesource.com/4259 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
2017-04-17Add unit test for CFX_SAXReader.Tom Sepez
Make the pending re-write CL smaller, and back-fill some test. Virtualize the handler interface so that it can be mocked. Change-Id: Icbbc6b6789ba135b73c6fd1bbb90f5aae6ab7649 Reviewed-on: https://pdfium-review.googlesource.com/4250 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
2017-04-13Add ability to make StringCs from std::vector<>s.Tom Sepez
Pre-cursor to using more std::vector<uint8_t> as byte buffers. The widestring test case is more complicated, where we're not sure of having any particular uint*_t type to match wchar_t. Change-Id: Ic27980f16cdbc61fac7c11f39a85eea58d19bacb Reviewed-on: https://pdfium-review.googlesource.com/4153 Reviewed-by: dsinclair <dsinclair@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
2017-04-13Use vector of unique ptr in cpdf_data_avail.cpp.Tom Sepez
Remove member m_pAcroForm as it is only used in one method, and immediately set upon entry. A local will be fine here. Change-Id: I373f68252249a248061828530098ad759907259a Reviewed-on: https://pdfium-review.googlesource.com/4190 Reviewed-by: dsinclair <dsinclair@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
2017-04-13Move another opertator<<() prototype into correct .h file.Tom Sepez
Prototype should be in header for .cpp file actually implementing it. Change-Id: I8f771a4840e09da419cee384e3d6dc2e7bba1cc4 Reviewed-on: https://pdfium-review.googlesource.com/4170 Commit-Queue: dsinclair <dsinclair@chromium.org> Reviewed-by: dsinclair <dsinclair@chromium.org>
2017-04-13core: allow building against system zlibMiklos Vajna
In a way similar to the existing USE_SYSTEM_LIBJPEG. The default is of course still the DEPS-based zlib checkout. Change-Id: I649e75beb1a19c4cf7abda42fa10ebe5babfe946 Reviewed-on: https://pdfium-review.googlesource.com/4070 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Lei Zhang <thestig@chromium.org>
2017-04-13Refcount CPDF_CryptoHandlerchromium/3076chromium/3075chromium/3074chromium/3073chromium/3072Tom Sepez
Avoid tracking ownership via m_bLocalCryptoHandler. Also remove m_bEncryptCloned, as it is always false. Replace some methods with direct calls to underlying code. Change-Id: Ifa9d6f721c59d07e3b8e258f76832ca9f2ea0fc9 Reviewed-on: https://pdfium-review.googlesource.com/4111 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Lei Zhang <thestig@chromium.org>
2017-04-12Cache TypeFace instead of FontCary Clark
Skia tracks the font passed so it can use it later to draw. In XFA's case, the font is discarded after Skia sees it, but before it draws with it. Track the TypeFace instead, which remains available across the XFA lifetime. R=dsinclair@chromium.org Bug: 705193 Change-Id: I4fc9fee4a7d96ea25f242975f6c0d10941f5c549 Reviewed-on: https://pdfium-review.googlesource.com/4058 Reviewed-by: Nicolás Peña <npm@chromium.org> Commit-Queue: Nicolás Peña <npm@chromium.org>
2017-04-12Some fixes to the fallback font code.Nicolas Pena
This CL applies several fixes to the fallback font code. - PDFium uses -1 to indicate that no glyph index was found, but freetype uses 0. In CPDF_TrueTypeFont, an index of 0 indicates a freetype failure, which means we should try to find the glyph from a fallback font. - Improve the fallback glyph calculation by going from original font charcode to unicode to fallback font charcode. - Consider the m_ExtGID on Mac when deciding the fallback. Bug: chromium:665467 Change-Id: I2be34983e0d768d9a598043f84edd2d70f033c86 Reviewed-on: https://pdfium-review.googlesource.com/4055 Commit-Queue: Nicolás Peña <npm@chromium.org> Reviewed-by: dsinclair <dsinclair@chromium.org> Reviewed-by: Tom Sepez <tsepez@chromium.org>
2017-04-12No gradient to draw if clip is emptyCary Clark
If a clip with bounds (0, 0, 0, 0) is used to specify the limits of a gradient, the clipping rectangle will be degenerate. Leave the clip empty in this case. R=dsinclair@chromium.org Bug: 705736 Change-Id: I4a5845459c1527f2c643563d6bef4d6dcc6781d3 Reviewed-on: https://pdfium-review.googlesource.com/4071 Reviewed-by: dsinclair <dsinclair@chromium.org> Commit-Queue: Cary Clark <caryclark@google.com>
2017-04-12Clip stack management was off by one.Cary Clark
To help find future bugs, add printf and checking to see that the skia clip stack and the pdfium clip stack are in sync. Bug: 705783,795821 Change-Id: I3ed39cbe2514ab18e5bee6eb363cc2d0f042bff5 Reviewed-on: https://pdfium-review.googlesource.com/4090 Reviewed-by: dsinclair <dsinclair@chromium.org> Commit-Queue: dsinclair <dsinclair@chromium.org>
2017-04-11Refactor CFX_Palettechromium/3069Nicolas Pena
The pointer members point to owned arrays. Avoid FX_Free in CFX_Palette. Use std::pair and avoid the use of a buggy quicksort. Change-Id: I5d5471d56dbfd32800e204c84664c436d1fbab08 Reviewed-on: https://pdfium-review.googlesource.com/4038 Commit-Queue: Nicolás Peña <npm@chromium.org> Reviewed-by: Tom Sepez <tsepez@chromium.org>
2017-04-10Guard against negative shift in nVal calculationDan Sinclair
It's possible for the nVal to become negative as it is shifted. This Cl changes nVal to be a checked_numeric and bails out if the shift is invalid. Bug: chromium:708504 Change-Id: Ia2ebbc828ece7f7d443432542784b39defe6a897 Reviewed-on: https://pdfium-review.googlesource.com/4010 Commit-Queue: dsinclair <dsinclair@chromium.org> Reviewed-by: Nicolás Peña <npm@chromium.org>
2017-04-10Use checked_numeric to guard shiftDan Sinclair
It's possible for the RANGELEN[NTEMP] value to be larger then 32. This will make the shift invalid if the 1 is an int. This CL changes to 1L and uses the CheckedNumeric to validate that the result is inside the needed range for an int. Bug: chromium:708439 Change-Id: I1f0359985c2d7769367bd0edcf5e081f5bb58816 Reviewed-on: https://pdfium-review.googlesource.com/3991 Commit-Queue: dsinclair <dsinclair@chromium.org> Reviewed-by: Nicolás Peña <npm@chromium.org>
2017-04-10Guard against negative shift in jbig2 huffman initializationDan Sinclair
Depending on the code table, it's possible to have the largest PREFLEN value in the huffman table to be > 32. This will, potentially, cause the calcuation of ((FIRSTCODE[i - 1] + LENCOUNT[i - 1]) << 1 to overflow the int value and cause a negative shift. This Cl checks the shift value and failes the initialization if we would shift a negative value. Bug: chromium:709781 Change-Id: Ia165a01ba9412e31c5e5a43717d415fcb42eafe5 Reviewed-on: https://pdfium-review.googlesource.com/3990 Reviewed-by: Lei Zhang <thestig@chromium.org> Reviewed-by: Nicolás Peña <npm@chromium.org> Commit-Queue: dsinclair <dsinclair@chromium.org>
2017-04-10Fix bytestring passing conventions, part 2.Tom Sepez
Change pass by reference to const reference or pointer. Change-Id: Ic007f14e6569679a846980a96cc627eac4ecd5d6 Reviewed-on: https://pdfium-review.googlesource.com/3953 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
2017-04-10Remove unreachable code in CFX_ScanlineCompositorchromium/3068Nicolas Pena
m_Transparency is set in the Init method of the CFX_ScanlineCompositor. After the removal of the icc transform, it is clear that m_Transparency can only take values strictly less than 64. Change-Id: I6c33b9971b3b8a1ec45fbf50901f9815f323a2a5 Reviewed-on: https://pdfium-review.googlesource.com/3970 Commit-Queue: dsinclair <dsinclair@chromium.org> Reviewed-by: dsinclair <dsinclair@chromium.org>
2017-04-08Add title (/T) extraction for PDF tagged structureschromium/3067chromium/3066dan sinclair
This CL adds the ability to extract the title from a tagged structure element if one exists. Bug: pdfium:672 Change-Id: I22e2a8371db4f08b8a70dd77002f1befab97f530 Reviewed-on: https://pdfium-review.googlesource.com/3819 Reviewed-by: Lei Zhang <thestig@chromium.org> Reviewed-by: Tom Sepez <tsepez@chromium.org> Commit-Queue: dsinclair <dsinclair@chromium.org>
2017-04-07Create initializer-list ctor for strings.Tom Sepez
Also use safe arithmetic for two-arg ctor. Change-Id: I5d541d9b2d5fe5b939f4cc8c22cf034f5cb01176 Reviewed-on: https://pdfium-review.googlesource.com/3955 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
2017-04-07Cleanup string passing in core/fpdf*Tom Sepez
Return strings where possible. Add missing consts to strings passed by ref. Convert non-const cases to pointers. Rename a few methods to be clearer. Change-Id: I86569bc1744f539e6dd67fc73649b272c016328c Reviewed-on: https://pdfium-review.googlesource.com/3951 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Nicolás Peña <npm@chromium.org>
2017-04-07Tweak CFDF_Font::AppendChar()Tom Sepez
Pass in/out argument as a pointer. Avoid pointless malloc just to copy in multibyte case. Then we can avoid special-casing the single-byte case. Change-Id: I3dd2d57e08ef6ad7b78ea38398b228fa41a9b3e6 Reviewed-on: https://pdfium-review.googlesource.com/3950 Reviewed-by: Nicolás Peña <npm@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
2017-04-07Remove the ICC transform where it is never initialized.Nicolas Pena
- All callers of CFX_DIBitmap methods use nullptr as the ICC Transform. - The CFX_ScanlineCompositor is always initialized with a null ICC transform. Many of its methods call TranslateScanline from the ICC module. This method casts just uses ((CLcmsCmm*)pTransform)->m_hTransform, and this should not be done when pTransform is nullptr. Change-Id: I1b846d1f20699fa017cb9a276be3214eb8fabd4b Reviewed-on: https://pdfium-review.googlesource.com/3931 Commit-Queue: Nicolás Peña <npm@chromium.org> Reviewed-by: Tom Sepez <tsepez@chromium.org>
2017-04-07Remove the "old way" of performing FlateUncompress.Tom Sepez
Repeated looping over Realloc() with a linearly increasing buffer size is going to be O(N^2) which is why it was limited to small cases. But it's wrong to do it anyways, and it kicks out a lurking issue in PartitionAlloc as it turns out. See https://crbug.com/709271, but this doesn't actually fix that issue, it just avoids it. Change-Id: I39d8f8df0d5fc358b7b9caa15c507a8f57e45004 Reviewed-on: https://pdfium-review.googlesource.com/3910 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
2017-04-07Use CFX_MaybeOwned on the buffer of CFX_DIBitmapNicolas Pena
Change-Id: Ia4181a2838de60e4e177165b3689338c86f45771 Reviewed-on: https://pdfium-review.googlesource.com/3817 Reviewed-by: Tom Sepez <tsepez@chromium.org> Commit-Queue: Nicolás Peña <npm@chromium.org>
2017-04-06Fix unit tests to initialize PartitionAlloc.Chris Palmer
Tests need to be subclasses of FPDF_Test. BUG=pdfium:700 Change-Id: I317ec2c49567e58cb57c6222e387574226f594b3 Reviewed-on: https://pdfium-review.googlesource.com/3890 Reviewed-by: Tom Sepez <tsepez@chromium.org> Commit-Queue: Chris Palmer <palmer@chromium.org>
2017-04-06Cleanup CFX_BitmapComposerNicolas Pena
Remove default params, use more unique_ptr, and fix some nits. Change-Id: I4ed9a576fb3ebd9afd64c6544d1f6a1103e9d270 Reviewed-on: https://pdfium-review.googlesource.com/3872 Reviewed-by: dsinclair <dsinclair@chromium.org> Commit-Queue: Nicolás Peña <npm@chromium.org>
2017-04-06Cleanup nits from previous XML Cldan sinclair
Change-Id: I1e0b84a8b208d92aff3f02b6461b3a20f60f4d1e Reviewed-on: https://pdfium-review.googlesource.com/3870 Reviewed-by: Nicolás Peña <npm@chromium.org> Commit-Queue: Nicolás Peña <npm@chromium.org>
2017-04-06Cleanup tagged code.dan sinclair
This Cl removes a bunch of unused tagged code. Some of this will need to come back in the future (like the attr code) but we can add it back with tests when needed. Bug: pdfium:672 Change-Id: I7aaed79963910b336f42ce665790408038c39ba4 Reviewed-on: https://pdfium-review.googlesource.com/3830 Reviewed-by: Tom Sepez <tsepez@chromium.org> Reviewed-by: Nicolás Peña <npm@chromium.org> Commit-Queue: dsinclair <dsinclair@chromium.org>
2017-04-06Cleanup the tagged codedan sinclair
This CL removes the IPDF_Struct* classes in favour of their only implementation. The tagged code was split out into files matching the classes they contain. The friendship between CPDF_StructTree and CPDF_StructElement was broken in favour of accessors. Bug: pdfium:672 Change-Id: Iade83b608fb7168b3b0f41338d10d5fd8ab91a6e Reviewed-on: https://pdfium-review.googlesource.com/3820 Reviewed-by: Tom Sepez <tsepez@chromium.org> Reviewed-by: Nicolás Peña <npm@chromium.org> Commit-Queue: dsinclair <dsinclair@chromium.org>
2017-04-06Remove the length reference from XMLSyntaxParserdan sinclair
This Cl removes the reference to the length of the CFX_BlockBuffer that was stored and manipulated in CFDE_XMLSyntaxParser. Methods have been added to BlockBuffer to satisify the usages in the syntax parser. Change-Id: I1107c343ce267283c4c45aa3ae1bbfa93c24079f Reviewed-on: https://pdfium-review.googlesource.com/3816 Commit-Queue: dsinclair <dsinclair@chromium.org> Reviewed-by: Nicolás Peña <npm@chromium.org>
2017-04-05Use unique_ptr in the children of CFieldTree::Nodechromium/3064Nicolas Pena
Change-Id: I149718bbdb7195223012150d6162d73cbeb3b8cc Reviewed-on: https://pdfium-review.googlesource.com/3813 Reviewed-by: Tom Sepez <tsepez@chromium.org> Reviewed-by: dsinclair <dsinclair@chromium.org> Commit-Queue: Nicolás Peña <npm@chromium.org>
2017-04-05use freetype if windows font decode failsCary Clark
If windows is unable to resolve the font stream, use the custom empty handler instead, which maps to freetype. R=dsinclair@chromium.org,bungeman@google.com Bug:chrome:705580 Change-Id: Ib1732f29f462fc5d5e42329088c80c83abe3ea25 Reviewed-on: https://pdfium-review.googlesource.com/3771 Reviewed-by: dsinclair <dsinclair@chromium.org> Commit-Queue: Cary Clark <caryclark@google.com>
2017-04-05Fix assignment to len.dan sinclair
Sigh, messed up std::min conversion. TBR=npm@chromium.org Bug: pdfium:697 Change-Id: I0091f178b08fa537ebc3962d9349eb46d0aa8a80 Reviewed-on: https://pdfium-review.googlesource.com/3751 Reviewed-by: dsinclair <dsinclair@chromium.org> Commit-Queue: dsinclair <dsinclair@chromium.org>
2017-04-05Use correct length in guard checkdan sinclair
When fixing https://crbug.com/672177 we added a guard that we aren't reading off the end of the file. That guard used the file access Position(). This is the wrong value to compare against as our read position and the file access Position may be different. This CL updates the check to use the correct current file position. Bug: pdfium:697 Change-Id: I68a5eaed2f1f3d65422605f0a8474144cfa7d172 Reviewed-on: https://pdfium-review.googlesource.com/3711 Commit-Queue: dsinclair <dsinclair@chromium.org> Reviewed-by: Tom Sepez <tsepez@chromium.org>
2017-04-04Remove a couple checks for MakeUnique failures.Lei Zhang
Combine some common CCodec_Jbig2Module code. Change-Id: I9a046314bc0e9dddc9a8c1a06b37764e9f3cc4b6 Reviewed-on: https://pdfium-review.googlesource.com/3713 Commit-Queue: Lei Zhang <thestig@chromium.org> Reviewed-by: Nicolás Peña <npm@chromium.org>
2017-04-04Remove dib_int.hNicolas Pena
This CL removes dib_int.h by moving CStretchEngine to its own file, other classes to where they are used, and the remaining to fx_dib.h. Change-Id: Ie2d4bb39389737cd631f92b88000ea942608da21 Reviewed-on: https://pdfium-review.googlesource.com/3714 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Nicolás Peña <npm@chromium.org>
2017-04-04RefCount CPDF_StreamAcc all the time.Tom Sepez
Pass stream argument to constructor; it feels like a stream accessor should always be made from a stream rather than passing one in after the fact. Change-Id: Iaa46cb37677b81f0170f5d39bab76ad38ea4af44 Reviewed-on: https://pdfium-review.googlesource.com/3620 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
2017-04-04Use some uint64_t constants in fx_crypt_sha.cpp.Lei Zhang
Instead of storing the same constants as strings and converting them to uint64_t. Change-Id: Ib32b0f721303f917e197cf93999857a7ee745823 Reviewed-on: https://pdfium-review.googlesource.com/3621 Reviewed-by: Tom Sepez <tsepez@chromium.org> Commit-Queue: Lei Zhang <thestig@chromium.org>
2017-04-04Cleanup CFX_BlockBufferDan Sinclair
This CL cleans up default parameters, return values and bits of code formatting in the CFX_BlockBuffer code. Change-Id: Ie7fd7d86046f716d0786213735ad464eefc58dbb Reviewed-on: https://pdfium-review.googlesource.com/3670 Commit-Queue: dsinclair <dsinclair@chromium.org> Reviewed-by: Nicolás Peña <npm@chromium.org> Reviewed-by: Tom Sepez <tsepez@chromium.org>
2017-04-04Let CPDF_ImageRenderer own the CFX_ImageRendererNicolas Pena
This CL avoids some void* usage and removes CancelDIBits after making a CPDF_ImageRenderer own its CFX_ImageRenderer. Bug: pdfium:686 Change-Id: Ied205c57a858cc14d8e2c592db3444ed465b2796 Reviewed-on: https://pdfium-review.googlesource.com/3673 Reviewed-by: Tom Sepez <tsepez@chromium.org> Commit-Queue: Nicolás Peña <npm@chromium.org>
2017-04-04RefCount CPDF_IccProfile all the timeTom Sepez
Make the IccProfile track its stream so that it has a proper key with which to purge the docpagedata map. Change-Id: Ib05ebc1afb828f1f5e5df62a1a33a1bfdecf507d Reviewed-on: https://pdfium-review.googlesource.com/3619 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>