Chrome Version       : 64.0.3273.0
OS Version: OS X 10.12.6

Turns out I have bookmark title that triggers the warning that cites  http://crbug.com/724880  now. It's my old pal' the red telephone, from  Issue 775404 

[60849:775:1120/112040.769192:ERROR:render_text_harfbuzz.cc(724)]  TextRunHarfBuzz error, please report at  crbug.com/724880 : range: {6,10}, rtl: 0, level: '', script: 25, font: '.AppleSystemUIFont', glyph_count: 3, pos: 6, glyph_to_char: 0->7, 1->8, 2->9,
[60849:775:1120/112040.769294:INFO:render_text_harfbuzz.cc(1329)] fdsfd☎️fds

Just the red telephone "☎️" is enough to trigger the warning:

[61335:775:1120/112436.145678:ERROR:render_text_harfbuzz.cc(724)]  TextRunHarfBuzz error, please report at  crbug.com/724880 : range: {1,2}, rtl: 0, level: '', script: 1, font: '.AppleSystemUIFont', glyph_count: 0, pos: 1, glyph_to_char:
[61335:775:1120/112436.145718:INFO:render_text_harfbuzz.cc(1329)] ☎️

Pasting "fdsfd☎️fds" into the omnibox on a Views browser can also cause a DCHECK:

[60849:775:1120/112259.167096:FATAL:render_text_harfbuzz.cc(749)] Check failed: cluster_begin_x <= cluster_end_x (29.4219 vs. 20.6855)
0   libbase.dylib                       0x0000000108bcb26e base::debug::StackTrace::StackTrace(unsigned long) + 174
1   libbase.dylib                       0x0000000108bcb32d base::debug::StackTrace::StackTrace(unsigned long) + 29
2   libbase.dylib                       0x0000000108bc95dc base::debug::StackTrace::StackTrace() + 28
3   libbase.dylib                       0x0000000108c62a6f logging::LogMessage::~LogMessage() + 479
4   libbase.dylib                       0x0000000108c603e5 logging::LogMessage::~LogMessage() + 21
5   libgfx.dylib                        0x000000010b2b8a52 gfx::internal::TextRunHarfBuzz::GetGraphemeBounds(gfx::RenderTextHarfBuzz*, unsigned long) const + 994
6   libgfx.dylib                        0x000000010b2b96ea gfx::internal::TextRunHarfBuzz::GetGraphemeSpanForCharRange(gfx::RenderTextHarfBuzz*, gfx::Range const&) const + 826
7   libgfx.dylib                        0x000000010b2bd8ff gfx::RenderTextHarfBuzz::GetCursorSpan(gfx::Range const&) + 735
8   libgfx.dylib                        0x000000010b2f64ef gfx::RenderText::GetCursorBounds(gfx::SelectionModel const&, bool) + 1231
9   libgfx.dylib                        0x000000010b2fc5a1 gfx::RenderText::UpdateCachedBoundsAndOffset() + 113
10  libgfx.dylib                        0x000000010b2fc519 gfx::RenderText::GetUpdatedCursorBounds() + 25
11  libviews.dylib                      0x0000000138bc2858 views::Textfield::UpdateCursorViewPosition() + 56
12  libviews.dylib                      0x0000000138bc206b views::Textfield::UpdateAfterChange(bool, bool) + 155
13  libviews.dylib                      0x0000000138bc1f93 views::Textfield::InsertOrReplaceText(std::__1::basic_string<unsigned short, base::string16_internals::string16_char_traits, std::__1::allocator<unsigned short> > const&) + 259


But just pasting the telephone doesn't cause a DCHECK.


There is related quirky behaviour around selection with this string when the DCHECK is ignored in Release (see attached)

Windows Canary seems fine with this string, and RenderTextMac renders the string OK (but of course can't select it).

So I think it's some combination of RenderTextHarfbuzz splitting the string incorrectly (i.e.  Issue 775404 ), and Harfbuzz+CoreText returning some bogus glyph positioning for the variation selector. Fun.


Diagnosing the warning, it's a failure in the `return false` at:

  Iterator element = std::upper_bound(elements_begin, elements_end, pos);
  if (element == elements_begin) {
    *chars = range;
    *glyphs = Range();
    return false;
  }

The glyph count for the orphaned variation selector is zero. I suppose this run "shouldn't exist", or it should render as tofu instead.

If there _is_ text after it, and something asks for the glyph/cluster for the variation selector, it can't be found.
 

Deleted: select_all.png 5.3 KB

	select_all.png 5.3 KB View Download

Deleted: Screen Shot 2017-11-20 at 11.32.38 am.png 9.9 KB

	Screen Shot 2017-11-20 at 11.32.38 am.png 9.9 KB View Download