Do not perform fallback for non-characters
In r492524 we resumed performing per-character fallback for unassigned characters. This was intended to avoid problems upgrading the emoji data file. When updating the emoji data before updating ICU new emoji may have appeared in the unassigned range, which we would erroneously exclude from fallback in that case. However, allowing the full range of unassigned characters does not make sense either and triggered a memory regression in our benchmark when the amazon.in page triggers fallback for the non printable U+FFFE sentinel character. To solve this, in addition to private use area codepoints we exclude 66 Unicode non-characters, defined in http://www.unicode.org/faq/private_use.html#nonchar4 They are part of the unassigned range, but have special meanings and are explicitly intended as non-characters. We must not try to look for glyphs in fonts for these. Bug: 862352 Change-Id: I50b22a4acb0d148595c5ae4f99a57d59ee8c41dd Reviewed-on: https://chromium-review.googlesource.com/1143277Reviewed-by:Koji Ishii <kojii@chromium.org> Reviewed-by:
Emil A Eklund <eae@chromium.org> Commit-Queue: Dominik Röttsches <drott@chromium.org> Cr-Commit-Position: refs/heads/master@{#576499}
Showing
Please register or sign in to comment