Repository navigation
Sync S8: merge upstream PDFium up to 3c9748bd6 (Sep 30 2026) - #71
Merged
Merged
Conversation
Change-Id: If68cbb8024de3140c26512dd6b3ae5ff97bd3233 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/156590 Commit-Queue: Lei Zhang <thestig@chromium.org> Auto-Submit: April Kallmeyer <ask@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: April Kallmeyer <ask@chromium.org>
The other changes are a copy of https://crrev.com/1676222 which updated a path and a condition used to obtain the ndk. Change-Id: I11bd1d242c5965038e004803c912790aa9f09311 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/156610 Auto-Submit: April Kallmeyer <ask@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: April Kallmeyer <ask@chromium.org>
Global Mean Squared Error (MSE) and per-pixel delta thresholds in image_diff can be overly restrictive on continuous-tone images (such as JPEG decompressed outputs or anti-aliased gradients). Subpixel inverse-DCT precision, chroma upsampling interpolation, and rounding variations between integer and floating-point decoders produce small delta clusters along high-contrast boundaries. While visually imperceptible, these local variations exceed strict per-pixel delta limits (delta <= 3). Conversely, simply raising the global MSE threshold allows localized defects in small page regions to go undetected. To resolve this, introduce a sliding window Mean Squared Error (CalculateMaxWindowMSE) using a 2D Summed-Area Table (Integral Image) with O(1) evaluation per pixel. This enforces that no local sub-window (default 8x8) exceeds a localized error density threshold, bounding local distortion while tolerating continuous-tone rounding noise. Extend the --fuzzy command-line flag and test suppressor to accept positional parameters: --fuzzy[=delta,mse,win_mse]. This change introduces the differ infrastructure and unit tests without altering existing test expectations or suppressions. TAG=agy CONV=1c2eb351-a6fc-4fe7-ab25-db1c71c96f52 Change-Id: Icd92b3cbd51616daa87f615240433e98086f198f Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/156190 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
Update EmbedderTest comparison helpers to accept DiffOptions instead of individual delta and MSE parameters as a mechanical refactoring without behavioral changes. -- Fold overload into a single form. TAG=agy CONV=2860a30e-1863-4ef5-b361-28d96ec73cb8 Change-Id: I9fc0b1f0903fdd1a2956cbd50c96ad608f4b40df Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/156650 Auto-Submit: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Lei Zhang <thestig@chromium.org>
Update CompareBGRxBitmapToPng() in testing/embedder_test.cpp to delegate to CalculatePixelsDifferent() and add unit tests covering exact and fuzzy matching. -- Make minor improvements in matching algorithm. TAG=agy CONV=2860a30e-1863-4ef5-b361-28d96ec73cb8 Change-Id: I5e20575bcc30bfc67757f54b73eafa27368b12af Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/156670 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
Wire CalculateMaxWindowMSE into EmbedderTest's bitmap comparison routines. When CompareBitmapWithFuzzyExpectationSuffix() is invoked, it now checks max_window_mean_squared_error using kMaxFuzzyWindowSize and kMaxFuzzyWindowMeanSquaredError from pixel_diff_util.h in addition to per-pixel delta and global MSE. This ensures embedder tests share the same localized defect rejection capabilities as pdfium_diff. This will catch more fuzzy failures. TAG=agy CONV=75945a90-5be1-4c24-a2c4-402f00166910 Change-Id: I9287d092d264a3e9d380002ba69ef7b06f1d1b8b Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/156630 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
With the pre-existing code, invalid 1-D runs can eventually cause an integer overflow. Check for invalid run length to return early and avoid the integer overflow altogether. Bug: 534431561 Change-Id: I0a5964ec41eab0d3b7c7741824a67fd76d0812f6 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/156450 Reviewed-by: Tom Sepez <tsepez@chromium.org> Commit-Queue: Lei Zhang <thestig@chromium.org>
This CL sets BROTLI_DECODER_PARAM_LARGE_WINDOW to 1 during decoder initialization. This allows for greater compliance with the BrotliDecode spec and allows for greater decode window sizes of up to 1GiB. Bug: 550470373 Change-Id: I795d1b363a2977d3b1e7a1d9b03ea03dd1a4ee3b Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/155710 Auto-Submit: Aryan Krishnan <aryankrishnan4b@gmail.com> Reviewed-by: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
Removes architectural bottlenecks for subsequent fontations work. Allow pdf_enable_fontations=true in standalone AGG builds without requiring pdf_use_skia=true. This enables safe Rust font table parsing via Fontations (read-fonts and skrifa) while preserving FreeType/AGG rasterization. Also add frankenfonts.md proposal document outlining the hybrid Rust font parsing and FreeType rasterization architecture. This CL does not accomplish the end goal, however. * Add --fontations flag support to pdfium_unittests TAG=agy CONV=db6cb163-190b-47d4-bbdc-b519626ce208 Bug: 42271123 Change-Id: I9abb1a76db071d984935ec47a65fd3431aacb0eb Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/156050 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
Remove `has_rust_decoders` from Suppressor to enforce no fuzzy blankets for Rust image decoders. Previously, the presence of any Rust decoder feature triggered blanket fuzzy diff matching across all pixel and corpus tests. With the introduction of windowed MSE and custom fuzzy tolerances in image_diff and SUPPRESSIONS_EXACT_MATCHING, lossless codecs (BMP, PNG, and GIF) are held to bit-perfect exact matching by default. Tests with acceptable decoder variances can now be selectively configured with custom tolerances instead of blanket forgiveness. TAG=agy CONV=9efbb0e0-217e-442a-9ecd-2e81b0992eb8 Change-Id: I8ca79f42b3326618955183e75861ea63da1b8ef6 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/156710 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
Also skip the test when running under AddressSanitizer in addition to debug builds, as it is slow as well. TAG=agy CONV=0541f91c-f7f9-4e30-91c4-8f5dcbe5c90c Change-Id: I62372265ceda1c4e00c92e1b9087714f29823202 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/156730 Reviewed-by: Lei Zhang <thestig@chromium.org> Auto-Submit: Tom Sepez <tsepez@chromium.org> Commit-Queue: Lei Zhang <thestig@chromium.org>
In horizontal mode, FaxG4GetRow() decodes two 1-D runs. Invalid runs can eventually cause an integer overflow when chained together. Bound the run lengths against the row width to return early and avoid the integer overflow altogether. Also check if FaxGetRun() returns a negative value before adding to the run length. This prevents adding -1 on unexpected EOF or invalid code words, which previously caused the inner loop to terminate early as if -1 were a terminating code. Remove the now-redundant post-loop checks for negative run lengths. TAG=agy CONV=93b9333d-4e34-48a0-b2ef-349e9f00b040 Change-Id: Ie50964caea573880449cb6b6c476b09530df4f8d Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/156470 Reviewed-by: Tom Sepez <tsepez@chromium.org> Commit-Queue: Lei Zhang <thestig@chromium.org>
A run length in CCITT Group 3 and Group 4 fax encoding is inherently non-negative and fits within uint16_t, as the maximum run in the lookup tables is 2560. Returning std::optional<uint16_t> replaces the -1 sentinel value with std::nullopt. Update callers in FaxG4GetRow() and FaxGet1DLine() to check for an empty optional rather than a negative integer, preventing accidental addition of -1 or treating -1 as a terminating code. TAG=agy CONV=93b9333d-4e34-48a0-b2ef-349e9f00b040 Change-Id: I4aa25905fb8189e22a97e5ff861f1483848b546b Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/156471 Commit-Queue: Lei Zhang <thestig@chromium.org> Reviewed-by: Tom Sepez <tsepez@chromium.org>
The 24-bit color scanlines provided to ProgressiveDecoder by both the BMP decoders and LibjpegJpegContext (which performs an in-place RGB2BGR conversion in ContinueDecode()) are arranged in BGR order in memory. Rename ProgressiveDecoderContextDelegate::Format::kRgb to Format::kBgr to accurately reflect that the buffer is in BGR byte order. This is a pure rename with zero behavior change, preparing for the introduction of true RGB decoding. Then do the same for the other formats so affected. TAG=agy CONV=0541f91c-f7f9-4e30-91c4-8f5dcbe5c90c Change-Id: Ic1290c836a0f27236666ea984ca3b3de1939b00f Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/156790 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
IsLeapYear() previously inverted the 400-year exception, incorrectly treating century years divisible by 400 (such as 2000) as common years and non-leap century years (such as 1900) as leap years. Correct the logic to treat years divisible by 400 as leap years and other century years as common years. Update test expectations for 11/12/2000 in ParseDateUsingFormatForValidMonthDay to account for the leap day. Add boundary test cases for 1900 and 2000 to GetYearFromTime and GetMonthFromTime, and add a GetDayFromTime test. TAG=agy CONV=217915ef-20b8-4b6e-849b-8360d32c7229 Fixed: 555821585 Change-Id: I3722f459134e25b060d4a08fb5969986ae543aaf Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/156832 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Andy Phan <andyphan@chromium.org>
CJS_Util::printd() supports formatting dates with full and abbreviated
day of week ("dddd" and "ddd"). However, it never populated tm_wday in
the struct tm used to format the string, leaving it at 0. As a result,
CJS_Util::printd() always printed "Sunday" or "Sun" for any date.
Add FX_GetDayOfWeekFromTime() to calculate the day of the week
(0 = Sunday ... 6 = Saturday) per ECMAScript specifications, and use it
to set tm_wday in CJS_Util::printd().
TAG=agy
CONV=217915ef-20b8-4b6e-849b-8360d32c7229
Fixed: 555848529
Change-Id: Ia42e3f4ae814475cf5625c697effbf625e7b4f8e
Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/156833
Reviewed-by: Lei Zhang <thestig@chromium.org>
Commit-Queue: Andy Phan <andyphan@chromium.org>
FX_ParseDateUsingFormat() previously added 2000 to all two-digit years in [0, 99], mapping every two-digit year into the 2000-2099 range (e.g., "85" became 2085 and "95" became 2095). Adobe Acrobat and pdf.js use a 50-year pivot (the Date Horizon rule) where two-digit years 00-49 are mapped to 2000-2049, and 50-99 are mapped to 1950-1999. Apply Acrobat's 50-year pivot to two-digit year parsing in FX_ParseDateUsingFormat(). TAG=agy CONV=217915ef-20b8-4b6e-849b-8360d32c7229 Fixed: 555821586 Change-Id: I37c63a5baad132673064c57c3f463778b9c4be6f Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/156850 Commit-Queue: Andy Phan <andyphan@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
Per ISO 32000-1 section 10.3 and Table 58, transfer functions (/TR) must map input color component values in [0.0, 1.0] to output values also in the range [0.0, 1.0]. While PDF functions generally (section 7.10) can declare a /Range outside [0.0, 1.0] for shading patterns or color spaces, transfer functions modulate device color channels directly. When populating the 256-entry lookup table in CreateTransferFunc() and LoadSMask(), unclamped float outputs were rounded and converted to uint8_t. Any function outputting values outside [0.0, 1.0] (such as a Type 4 PostScript calculator function with /Range [-1 1]) wrapped modulo 256, producing inverted colors (e.g. -0.5 wrapped to 128 instead of saturating at 0). Clamp the rounded sample values to [0, 255] before storing them in the lookup table. This does not affect general function evaluation in CPDF_Function::Call(), so functions requiring /Range outside [0.0, 1.0] for shadings and color spaces continue to evaluate without restriction. Also add a unit test covering out-of-range transfer functions and update the existing test expectations that asserted the unclamped wrapped values. TAG=agy CONV=6c24f844-e095-4a12-a6ae-19ea09a39475 Fixed: 555967331 Change-Id: I610f5e1869d3e6dc525e7010ac3429d1901dd308 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/156830 Commit-Queue: Andy Phan <andyphan@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
Per ISO 32000-1 Table 58, when the /TR transfer function entry in an ExtGState dictionary is an array of functions, the elements correspond to the process colorants in order: Red, Green, Blue, (Gray), or Cyan, Magenta, Yellow, Black. In CreateTransferFunc(), the array was loaded into pFuncs using pFuncs[2 - i] = Load(array[i]), loading array element 0 into pFuncs[2] and array element 2 into pFuncs[0]. Downstream, pFuncs[0] populates samples_r and pFuncs[2] populates samples_b. This transposed the Red and Blue channels. The reversal originated from confusing physical BGR DIB scanline layouts with logical channel indices, and existing unit test expectations were written to match the reversed behavior. Load the transfer function array in forward order so element 0 modulates Red and element 2 modulates Blue. Reorder the test sample definitions and update unit test expectations in CPDFDocRenderDataTest to reflect the corrected channel mappings. TAG=agy CONV=6c24f844-e095-4a12-a6ae-19ea09a39475 Fixed: 555940409 Change-Id: I0ea622db48f7bed0d61a445d015891f2d25c1b9f Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/156831 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Andy Phan <andyphan@chromium.org>
When pdf_enable_rust_jpeg is set, JPEG decoding previously attempted to delegate to Skia's SkJpegDecoder wrapper. However, Skia's wrapper required pdf_use_skia, lacked direct Rust JPEG FFI bindings, and failed for 4-component (CMYK/YCCK) JPEGs, causing decoding to fall back to the C++ libjpeg implementation. This change introduces RustScanlineDecoder and RustJpegContext, which directly invoke the zune-jpeg Rust crate via CXX bridge FFI (rust_jpeg_ffi). This enables native Rust JPEG decoding for 1, 3, and 4-component (CMYK and YCCK) images, and removes the dependency on pdf_use_skia for Rust JPEG decoding. The obsolete skia_jpeg_context and skia_scanline_decoder implementation files have been removed. Continuous-tone JPEG test tolerances under Skia and embedder test expectations have been updated to accommodate minor IDCT rounding and chroma upsampling differences in zune-jpeg. TAG=agy CONV=873eda5c-74cd-479c-b5f1-24997eba2b24 Change-Id: I7fe24d45c8558cb64c5bc6acc0128d0ce242c434 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/155570 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
Every neighbor is read through GetLeftValue() / GetUpValue() / GetUpperLeftValue(), each of which does a span bounds check plus an `i >= bytes_per_pixel` or `empty()` branch, per byte and per filter. Size the row views once up front and hoist the two special cases out of the loops: the first bytes_per_pixel bytes of a row, where the left and upper-left neighbors are zero, and the first row of the image, where up and upper-left are zero. On the first row the Paeth predictor always selects the left neighbor and on the lead bytes it always selects up, so those reduce to the Sub and Up filters. The Up filter walks its three rows with fxcrt::Zip(), as do the lead bytes of the Average and Paeth filters. The Sub filter gets dedicated loops for 3- and 4-byte pixels that walk the row as fixed-size pixel structs and carry each byte channel in a local. Its channels form independent additive chains, and the generic loop re-reads dest_span[i - bpp] a few iterations after storing it, which serializes on store-to-load forwarding. Bug: 549230311 Change-Id: I97629d99f412c2d93bf7961aec00a93bd4d90ac8 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/155510 Reviewed-by: Lei Zhang <thestig@chromium.org> Reviewed-by: Tom Sepez <tsepez@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
Clean up legacy Hungarian notation and single-letter variable names to follow the style guide. Also use pre-increments and const consistently. This is a pure refactoring with no behavioral changes. TAG=agy CONV=f96089b4-cb0a-43a3-8bf4-16d4c0790cb3 Bug: 555848528 Change-Id: Ic47035b5d3a4ac9b60e1907a8634bd3ee02ffc3f Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/156910 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Andy Phan <andyphan@chromium.org>
Replace cascading lookahead checks with direct token length counting:
- Previously, token lengths 1 to 4 were determined by cascading
lookahead conditions with duplicated switch statements for each
length.
- Count consecutive identical format characters up front to determine
token length directly.
- Group lengths 1 and 2 together, since month, day, hour, minute, and
second tokens share identical up-to-2-digit integer parsing logic.
Only year ('y' vs 'yy') and AM/PM ('t' vs 'tt') require length checks.
This is a pure refactoring with no behavioral changes.
TAG=agy
CONV=f96089b4-cb0a-43a3-8bf4-16d4c0790cb3
Bug: 555848528
Change-Id: I5ae5017f5a90829e55ccba5579ab7c29f0a9aba8
Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/156911
Reviewed-by: Lei Zhang <thestig@chromium.org>
Commit-Queue: Andy Phan <andyphan@chromium.org>
The motivation is that I'd like to optimize the huffman symbol reading loop that's currently repeated in 1. CJBig2_HuffmanDecoder::DecodeAValue() 2. CJBig2_Context::DecodeSymbolIDHuffmanTable() (for reading `runcode`) 3. CJBig2_TRDProc::DecodeHuffman() (for reading `IDI`) I mostly care about optimizing the last one, but it seems nicer to dedupe them first. This implements the 2nd in terms of the 1st; the 3rd is for a future CL. * ParseFromStandardTable() becomes ParseFromTable() and receives a HuffmanTable instead of an index. This simplifies the CJBig2_HuffmanTable(int) constructor a bit, which used to load each built-in table in the ctor, and then again in ParseFromStandardTable(). This requires moving `struct HuffmanTable` to the header. (It's still an internal type, but as it's used by private functions, we have to move it over. If we wanted, we could put it inside the class, in the `private:` section.) * Add a constructor that builds a CJBig2_HuffmanTable from just code lengths. * Call that from CJBig2_Context::DecodeSymbolIDHuffmanTable() While here, add a comment over `lines`, and fix `RANDELEN` -> `RANGELEN` typo. No intended behavior change. (Linked bug vaguely related: I want to bump some limit, but things are currently very slow with high symbol count, even with the current limit.) Bug: 555145149 Change-Id: If29d5d11951a47cf7b32f61d6147054f3fffd684 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/156810 Commit-Queue: Nico Weber <thakis@chromium.org> Reviewed-by: Nico Weber <thakis@google.com> Reviewed-by: Tom Sepez <tsepez@chromium.org>
https://chromium.googlesource.com/chromium/src/third_party/zlib.git/+log/c5cc9edf8992..285e94b8fa95 If this roll has caused a breakage, revert this CL and stop the roller using the controls here: https://autoroll.skia.org/r/zlib-pdfium Please CC akall@google.com,thestig@chromium.org on the revert to ensure that a human is aware of the problem. To report a problem with the AutoRoller itself, please file a bug: https://issues.skia.org/issues/new?component=1389291&template=1850622 Documentation for the AutoRoller is here: https://skia.googlesource.com/buildbot/+doc/main/autoroll/README.md Bug: None Tbr: akall@google.com Change-Id: I92895061d92445e2d0da2f2cb23c712efe552eba Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/157090 Bot-Commit: pdfium-autoroll@skia-public.iam.gserviceaccount.com <pdfium-autoroll@skia-public.iam.gserviceaccount.com> Commit-Queue: pdfium-autoroll@skia-public.iam.gserviceaccount.com <pdfium-autoroll@skia-public.iam.gserviceaccount.com>
Passive FFI callbacks (e.g. FFI_OutputSelectedRect, FFI_Invalidate) are notifications invoked during internal operations such as painting. Embedders must not re-entrantly execute form actions or load/close pages during these callbacks. Track passive callbacks using CPDFSDK_FormFillEnvironment's in_passive_ffi_callback_ flag, and early-return from entry points such as FORM_OnAfterLoadPage, FORM_OnBeforeClosePage, and FORM_Do*Action when invoked re-entrantly. Active retrieval callbacks such as FFI_GetPage remain unflagged, as page loading is expected in those paths. TAG=agy CONV=9e87637a-cb01-4106-9c04-ad4da437fc61 Bug: 556535630 Change-Id: Id892141ce082cc8895b4ed5dc66af3edaf5ef740 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/156913 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
These checks were temporary debugging scaffolding to verify parity between FreeType and Skia/Fontations, but are now obsolete and divergent on edge cases. Eliminating them simplifies CFX_Face methods to direct return statements and removes unused Skia header inclusions. Change-Id: I6d03431be951ee856bd2dcde3dcfdb0d115a3183 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/156970 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
In opj_j2k_copy_default_tcp_and_create_tcd(), copy m_nb_mcc_records from the default tile coding parameters along with m_nb_max_mcc_records. TAG=agy CONV=a534277b-a66b-488e-ab0a-8833cb8e63ea Bug: 541707261 Change-Id: Ic9b58719dd662e067751e141782d629e93149bc6 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/157011 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
…ions" This reverts commit a0a39c3. Reason for revert: Reverted too much code. Original change's description: > Remove PDF_ENABLE_SKIA_TYPEFACE_CHECKS and dead parity assertions > > These checks were temporary debugging scaffolding to verify parity > between FreeType and Skia/Fontations, but are now obsolete and > divergent on edge cases. Eliminating them simplifies CFX_Face methods > to direct return statements and removes unused Skia header inclusions. > > Change-Id: I6d03431be951ee856bd2dcde3dcfdb0d115a3183 > Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/156970 > Reviewed-by: Lei Zhang <thestig@chromium.org> > Commit-Queue: Tom Sepez <tsepez@chromium.org> TBR=thestig@chromium.org,tsepez@chromium.org,pdfium-scoped@luci-project-accounts.iam.gserviceaccount.com No-Presubmit: true No-Tree-Checks: true No-Try: true Change-Id: I587788e995165d887e7d9c95aadb3976102c79c3 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/157111 Commit-Queue: Tom Sepez <tsepez@chromium.org> Bot-Commit: rubber-stamper@appspot.gserviceaccount.com <rubber-stamper@appspot.gserviceaccount.com>
Avoid ignoring fatal errors during Windows JPEG compression by using longjmp to catch failures in libjpeg C wrapper functions, matching the decompression architecture. -- use modern memory ownership model. -- avoid array of [1]. TAG=agy CONV=2e125a78-6471-43e2-9eed-d2106f60cbec Bug: 556853443 Change-Id: I2e5eadc41d277f2098ebc91be91d67578c10301f Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/157030 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
Allow CFX_Face creation without FreeType when using the Fontations backend for Type 1 fonts, skipping FT_FaceRec allocation and initialization. Handle Type 1 character mapping, encoding selection, glyph index lookup, and glyph path generation directly via Skrifa when FT_FaceRec is null, while preserving FT_CharMap handling for SFNT faces that still retain an FT_FaceRec. Guard all remaining FT_FaceRec accesses in CFX_Face against null and add unit test coverage for faceless Type 1 glyph paths and bitmap rasterization. -- Move Adobe Glyph List name and unicode conversions into cfx_face.cpp -- Suppress one test which will fail under the new mode until the follow-on CL implements LCD-like filtering. TAG=agy CONV=bfb4bf1b-47cc-4055-aafa-26a5312d81ea Bug: 42271123 Change-Id: I947c6f0734e42fe8a3e7d31c0cffa4b04469489a Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/158611 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
When rasterizing glyphs via `FT_Outline_Get_Bitmap()` in the stateless Fontations path (where `FT_FaceRec` is null), the code does not apply the LCD bounding-box padding or the 5-tap subpixel filter that `FT_Render_Glyph()` normally performs for `FT_RENDER_MODE_LCD`. Expand the horizontal bounding box by 2/3 of a pixel (128 units in 3x 26.6 fixed-point) on each side before rasterization and apply the 5-tap `FT_LCD_FILTER_DEFAULT` weights across each output scanline afterwards. Adjust the example_065.pdf fontations entry in testing/SUPPRESSION, better rounding in the new code has moved an edge. Bug: 42271123 Change-Id: Id4f30ab5c49413139efd0967ed8dc4e96c640a94 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/158810 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
Add CFXFontMapperSystemFontInfoTest.GetCachedTTCFaceWithLargeTtc to test CFX_FontMapper::GetCachedTTCFace() when the TTC data size exceeds the 1024-byte buffer in CFX_FontMapper::GetChecksumFromTT(). Disable the test on MSAN builds for now, as GetChecksumFromTT() reads from an uninitialized buffer when GetFontData() does not fill it. TAG=agy CONV=059f2a05-d124-4d35-b1e3-144227fa7db5 Bug: 565837689 Change-Id: Ia1e7823c663dd2058f777003624a7016f6be6f96 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/158730 Reviewed-by: Andy Phan <andyphan@chromium.org> Commit-Queue: Lei Zhang <thestig@chromium.org>
Previously, CFX_FontMapper::GetChecksumFromTT() passed an uninitialized 1024-byte stack buffer to SystemFontInfoIface::GetFontData(). When the TTC data exceeded 1024 bytes, GetFontData() returned the required size without writing to the buffer, causing GetChecksumFromTT() to compute a checksum over uninitialized memory. Fix this issue by reading the full `ttc_size` bytes once in CFX_FontMapper::GetCachedTTCFace() and passing the resulting span to GetChecksumFromTT(). This fixes the uninitialized memory read and reduces the number of GetFontData() calls from 2 to 1. Since GetChecksumFromTT() no longer calls GetFontData(), it can now live in the anonymous namespace. With the issue fixed, re-enable CFXFontMapperSystemFontInfoTest.GetCachedTTCFaceWithLargeTtc in MSAN builds. Now that GetFontData() only gets called once, also adjust other existing test cases to expect fewer GetFontData() calls. TAG=agy CONV=059f2a05-d124-4d35-b1e3-144227fa7db5 Bug: 565837689 Change-Id: Ifbe40ae1a1607d88d4f6099017297751d815ff65 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/158731 Reviewed-by: Andy Phan <andyphan@chromium.org> Commit-Queue: Lei Zhang <thestig@chromium.org>
When multiple fuzzy lines in testing/SUPPRESSIONS match the same test, Suppressor._LoadSuppressions() previously overwrote the entry in exact_matching_suppression_dict on each match, causing the last matching line in the file to win regardless of how many wildcards it contained. Score each matching fuzzy rule's predicate columns (columns 1 through 5) as a little-endian bitmask where '*' is 0 and any specific value is 1, retaining the highest-scoring rule. Re-order one line in testing/SUPPRESSIONS which failed under the old order-of-appearance rules. -- Narrow the scope of two other suppressions based upon the find_surprise_passes.py script. Change-Id: Ic35aec79fe64294b37ed632b1fa4432b678d87a7 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/158830 Commit-Queue: Lei Zhang <thestig@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
https://chromium.googlesource.com/chromium/src/tools/code_coverage.git/+log/e1878fbb1ff7..a81032f8d6f5 If this roll has caused a breakage, revert this CL and set the roller to dry run mode using the controls here: https://autoroll.skia.org/r/code-coverage-pdfium Please CC akall@google.com,dhoss@chromium.org,thestig@chromium.org on the revert to ensure that a human is aware of the problem. To report a problem with the AutoRoller itself, please file a bug: https://issues.skia.org/issues/new?component=1389291&template=1850622 Documentation for the AutoRoller is here: https://skia.googlesource.com/buildbot/+doc/main/autoroll/README.md Bug: None Tbr: akall@google.com Change-Id: I7056991f2a1e36bcbf56b25a4afad21a47afe301 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/158931 Commit-Queue: pdfium-autoroll@skia-public.iam.gserviceaccount.com <pdfium-autoroll@skia-public.iam.gserviceaccount.com> Bot-Commit: pdfium-autoroll@skia-public.iam.gserviceaccount.com <pdfium-autoroll@skia-public.iam.gserviceaccount.com>
https://chromium.googlesource.com/catapult.git/+log/50c971fc0958..7cce2c2359d9 If this roll has caused a breakage, revert this CL and set the roller to dry run mode using the controls here: https://autoroll.skia.org/r/catapult-pdfium Please CC akall@google.com,dhoss@chromium.org,thestig@chromium.org on the revert to ensure that a human is aware of the problem. To report a problem with the AutoRoller itself, please file a bug: https://issues.skia.org/issues/new?component=1389291&template=1850622 Documentation for the AutoRoller is here: https://skia.googlesource.com/buildbot/+doc/main/autoroll/README.md Bug: None Tbr: akall@google.com Change-Id: If60f21ff411ef846be1eb87da01953b1e867b167 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/158930 Commit-Queue: pdfium-autoroll@skia-public.iam.gserviceaccount.com <pdfium-autoroll@skia-public.iam.gserviceaccount.com> Bot-Commit: pdfium-autoroll@skia-public.iam.gserviceaccount.com <pdfium-autoroll@skia-public.iam.gserviceaccount.com>
This reverts commit db1b784. Reason for revert: Culprit for tree closure Failure Link: https://ci.chromium.org/ui/p/pdfium/builders/ci/win_xfa_asan/10568 Original change's description: > Roll Catapult from 50c971fc0958 to 7cce2c2359d9 (12 revisions) > > https://chromium.googlesource.com/catapult.git/+log/50c971fc0958..7cce2c2359d9 > > If this roll has caused a breakage, revert this CL and set the roller > to dry run mode using the controls here: > https://autoroll.skia.org/r/catapult-pdfium > Please CC akall@google.com,dhoss@chromium.org,thestig@chromium.org on the revert to ensure that a human > is aware of the problem. > > To report a problem with the AutoRoller itself, please file a bug: > https://issues.skia.org/issues/new?component=1389291&template=1850622 > > Documentation for the AutoRoller is here: > https://skia.googlesource.com/buildbot/+doc/main/autoroll/README.md > > Bug: None > Tbr: akall@google.com > Change-Id: If60f21ff411ef846be1eb87da01953b1e867b167 > Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/158930 > Commit-Queue: pdfium-autoroll@skia-public.iam.gserviceaccount.com <pdfium-autoroll@skia-public.iam.gserviceaccount.com> > Bot-Commit: pdfium-autoroll@skia-public.iam.gserviceaccount.com <pdfium-autoroll@skia-public.iam.gserviceaccount.com> TBR=pdfium-autoroll@skia-public.iam.gserviceaccount.com,pdfium-scoped@luci-project-accounts.iam.gserviceaccount.com,akall@google.com No-Presubmit: true No-Tree-Checks: true No-Try: true Bug: None Change-Id: I34ca5b6e1230bbe068d5d9f7498d6ddc8dc49418 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/158970 Reviewed-by: Tom Sepez <tsepez@chromium.org> Bot-Commit: rubber-stamper@appspot.gserviceaccount.com <rubber-stamper@appspot.gserviceaccount.com> Commit-Queue: Aryan Krishnan <aryankrishnan4b@gmail.com>
Instead of iteratively calling PutArrayElementReentrant() on an empty array, accumulate property names into a v8::LocalVector and instantiate the v8::Array in a single call to v8::Array::New(). Bug: 566111382 Change-Id: I55f4718719cfa613f2c50a5dff10222b95a96c11 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/158890 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
This was the only location that included dragonbox.h, and the '#if' style is similar to https://pdfium-review.googlesource.com/c/pdfium/+/154490. Change-Id: I58580a9283c64dba135f19af9a9325c04831e3af Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/158950 Reviewed-by: Tom Sepez <tsepez@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
Adjust expected LCD glyph width in TestType1Face from 18 to 24 to account for 5-tap LCD filter padding. This test currently early-exits on CQ because bots do not pass --fontations when running unittests, but needs to be corrected before bots run unittests with the Fontations backend. Change-Id: I1505af87d046b34b502154a368052f1938a33c1e Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/159030 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
Help to decouple FT and Fontations during tests. Remove cross-backend FreeType usage from Skrifa unit tests. Compare bounding boxes and font metadata directly against known constants for static test fonts instead of querying FreeType at runtime. Also remove GetFTFaceForTesting() from CFX_Face since this was its sole consumer. Change-Id: I7cbfb0e67a4ae75542dbd4f5c539fccc53070234 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/159050 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
Migrate the legacy .vpython3 spec to vpython.toml and vpython.toml.uv.lock, configure vpython_spec_files in build_overrides/build.gni, and update testing/tools/BUILD.gn data to reference the new specs. Bug: 491263752 LSC: go/vpython-uv-config-lsc TAG=agy CONV=b53bdf70-37e7-4696-be50-8ee6fc19b1f8 Change-Id: Iadf2a05a373d031d33a5fdacf872b9d8fdd05038 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/158350 Commit-Queue: Lei Zhang <thestig@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
Add Fontations implementations for CFX_Face::IsTtOt(), IsItalic(), IsBold(), and GetBBox() when running in Fontations-only or minimal FreeType configurations. In CFX_Face::GetBBox(), invert the FT_FaceRec check so that the dummy bounding box fallback remains at the end of the method. Expose is_sfnt(), is_italic(), is_bold(), and font_bbox() across the Skrifa CXX bridge to provide the underlying metrics from SFNT, Type 1, and CFF font formats. Bug: 42271123 Change-Id: Ie706c0d5961e819b820c4c23ac704bbdaff90065 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/158990 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
Add an IsFontations() implementation to CFX_Face::GetCharBBox(). This computes bounding boxes via GetGlyphBBox() and clamps them to the face ascender and descender when IsTricky() is true, avoiding a null pointer dereference on FT_FaceRec in Fontations-only mode. Also ensure the FreeType fallback returns early if the face record is null. Bug: 42271123 Change-Id: Ic7693d598a5fd6f58ac7d97bfe8b5c09fb50f6a0 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/159110 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
In a future where embedders may be enforcing the same unsafe checks as in a standalone build, be sure to trigger a rebuild of pdfium-using components if the unsafe buffers file changes. Due to the infrequency with which this file change, land this without worry of triggering extra chromium builds. Bug: 565766512 Change-Id: I4557e50971a97622fd448b8afb021f0515906351 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/159070 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
npm@ is leaving, so remove them from the root OWNERS file and replace TODO(npm) with TODO(crbug.com/...) pointing at newly filed bugs. No functional changes. Bug: 567626894, 567623562, 567621440, 567622442, 567626895, 567627774, 567626109 Change-Id: Id4f926858adf7da7e8f64013ed0488d96fa1cfe8 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/159150 Reviewed-by: Lei Zhang <thestig@chromium.org> Reviewed-by: Tom Sepez <tsepez@chromium.org> Commit-Queue: Nicolás Peña <npm@chromium.org>
Remove / replace inactive owner. Change-Id: I1254ede1a1a0f1cfe996486cd0a9683ab36a4fe1 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/148894 Reviewed-by: Dominik Röttsches <drott@chromium.org> Commit-Queue: Dominik Röttsches <drott@chromium.org>
Add an IsFontations() implementation to CFX_Face::GetSfntTable(). Expose a get_sfnt_table() method on skrifa::SkrifaFont via the CXX bridge that retrieves SFNT table data using FontRef::table_data() from read-fonts. When an empty buffer is passed, the method returns the length of the SFNT table (or 0 if missing/not an SFNT font). When a buffer is provided, it copies the table bytes and returns the table length. Bug: 42271123 Change-Id: I57170c972666fcf2f4a6f3bd348dd985d8b562dd Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/159130 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
This CL updates pdfium_test.cc to enable Brotli by default in the test application. This CL also replaces the --enable-brotli test flag with --disable-brotli due to it being enabled by default. Bug: 475855993 Change-Id: I55ef3b5e1dda767c0ede9552fad6c83e94a42bfc Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/158010 Commit-Queue: Aryan Krishnan <aryankrishnan4b@gmail.com> Reviewed-by: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
When rendering text with large transformation matrix entries (such as large c or d components that are not routed to DrawTextPath() by CFX_RenderDevice::DrawNormalText()), multiplying the matrix entries by 10000 and casting float to int via static_cast<int> in UniqueKeyGen can trigger float-cast-overflow undefined behavior. Use pdfium::saturated_cast<int> instead of static_cast<int> when computing the glyph cache key. PiperOrigin-RevId: 991122773 Change-Id: Ie63f9ad396b6c85b16575de7b858256b6ddfa4b5 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/159250 Commit-Queue: Lei Zhang <thestig@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
LibjpegScanlineDecoder previously used a const_cast in GetWritableSrcData() to mutate the input span in place when repairing missing trailers or corrupted scanner headers. Avoid mutating the caller's source span by implementing copy-on-write in GetWritableSrcData(). When repair is necessary, allocate and populate an owned DataVector buffer member called `corrected_src_data_`. -- Remove redundant DCHECK() give checked span usage. Fixed: 567106301 Change-Id: Icdefd8592b0600333a0b16ef9ecba3329310c365 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/159210 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
Introduce fxcrt::ToDataVector(range) in core/fxcrt/data_vector.h to construct a DataVector<T> from a range or span in C++20 without verbose iterator-pair construction until C++23 std::from_range is available. -- Remove redundant CHECK() give checked span usage. Change-Id: I2f16abbec60495e441fee13dbea64958613d43dd Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/159230 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
This CL adds the FPDF_StructElement_GetRawType() API alongside the backend API CPDFStructElement::GetRawType() to access the raw, unprocessed value of the /S entry. This allows embedders to be able to access original values if needed without editing the behaviour of existing, stable APIs that embedders may already use. Also update the CPDFStructElement class constructor to use this new API. Bug: 42270951 Change-Id: I55cdbe91a4c74fca7db16fc6adad7cd0dab75d15 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/158130 Reviewed-by: Andy Phan <andyphan@chromium.org> Commit-Queue: Andy Phan <andyphan@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
Add helper for ordinary std::vector<> aligned with what was added for PDFium's DataVector<>. Change-Id: I8546a212953a82aae44af62b3d315a43e738485d Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/159330 Commit-Queue: Tom Sepez <tsepez@chromium.org> Reviewed-by: Lei Zhang <thestig@chromium.org>
Convert a few instances missed in the prior CL. Change-Id: I73cfc023750d539b1394b29dec7bcfaed9149208 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/159310 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
Avoid mutating the input span in-place with const_cast in RustJpegScanlineDecoder. Make a copy of the input source data into a vector on demand when patching is required, matching the behavior of LibjpegScanlineDecoder. Change-Id: I0ead4a23f95ce24518ebec1da07e52e279bc9302 Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/159270 Reviewed-by: Lei Zhang <thestig@chromium.org> Commit-Queue: Tom Sepez <tsepez@chromium.org>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
12 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.