Regression:English translation of content doesn't appear properly after loading the page on Mac OS.
Reported by
vku...@etouch.net,
Jun 26 2017
|
||||||||||
Issue descriptionChrome Version: 61.0.3141.0 (Official Build) Revision 180095eb1bca7df1cdcb02547340499c2ee3af6e-refs/heads/master@{#482153} (32/64-bit) OS:Windows(7,8,10), Mac (10.11.6, 10.12.1), Linux (14.04 LTS) What steps will reproduce the problem? (1)Launch chrome and navigate to http://www.stepupnihongo.com/images/SUN3_L-51.pdf (2)Scroll down the page and observe the content of page Actual: English translation of content doesn't appear properly after loading the page. Expected: English translation of content should be properly displayed. This is a regression issue broken in 'M61' and below is the manual regression range Good Build: 61.0.3138.0 Bad Build: 61.0.3140.0
,
Jun 26 2017
Able to reproduce the issue and below are the bisect details obtained using per-revision bisect: You are probably looking for a change made after 481560 (known good), but no later than 481561 (first known bad). CHANGELOG URL: The script might not always return single CL as suspect as some perf builds might get missing due to failure. https://chromium.googlesource.com/chromium/src/+log/a95e0bf52bfe99622e4bf498adb761cf9eb218a9..1545ba94087279c773705760d51b728d709fd72d @dsinclair: Assigning to you, kindly take a look into it Please help us to find an owner if not with respect to your change. Adding Blocker label, please decrease or remove priority if not the case. Thanks.!
,
Jun 26 2017
,
Jun 26 2017
Looks like it picked a different font to use when rendering. Given the regression CL probably due to switching on Skia paths rendering.
,
Jun 26 2017
This is a recent regression but it also happens to be broken in Stable so ReleaseBlock-Dev is not appropriate even if we hand't already disabled Skia paths. FYI this suggests my spacing CL [1] does not seem to be working properly when Skia paths is enabled. [1] https://pdfium-review.googlesource.com/c/6630/
,
Jun 26 2017
CFX_SkiaDeviceDriver::DrawDeviceText seems to be ignoring the m_AdjustMatrix even when m_bGlyphAdjust is true, see for example https://cs.chromium.org/chromium/src/third_party/pdfium/core/fxge/ge/cfx_renderdevice.cpp?type=cs&l=951
,
Jun 26 2017
re #8 : thanks for the diagnosis -- very helpful
,
Jun 27 2017
dsinclair@ turned off Skia path for PDFium, so this is no longer a P1 regression.
,
Jun 27 2017
While this PDF does indeed use m_AdjustMatrix to condense characters, it mostly uses multiple master axis parameters to narrow the text in some places an embolden it in others. The multiple master control code is in cfx_font.cpp, CFX_Font::AdjustMMParams. I have a bit of rewriting to do to take advantage of this, and will work with our resident font expert to make sure that the result is sustainable.
,
Jun 30 2017
,
Jul 10 2017
|
||||||||||
►
Sign in to add a comment |
||||||||||
Comment 1 by vku...@etouch.net
, Jun 26 2017Summary: Regression:English translation of content doesn't appear properly after loading the page on Mac OS. (was: Regression:English translation of content doesn't appear properly after loading the page.)
321 KB
321 KB View Download
359 KB
359 KB View Download