Fix excessive combining marks causing GPU overload and rendering corruption

[vserediuk]

Fix excessive combining marks causing GPU overload and rendering corruption

Problem

Text containing a large number of Unicode combining characters (such as ⃝ U+20DD COMBINING ENCLOSING CIRCLE) stacked on a single base character causes:

1. GPU overload — the text renderer tries to draw hundreds/thousands of overlapping glyphs for a single character
2. Incorrect display — the rendered text appears as visual noise/garbage

Example of malicious text: ы҉⃝ repeated hundreds of times — a single Cyrillic letter with COMBINING CYRILLIC MILLIONS SIGN (U+0489, Mark_NonSpacing) followed by COMBINING ENCLOSING CIRCLE (U+20DD, Mark_Enclosing).

Root Cause

The IsDiacritic() function in ui/text/text.cpp only checked for QChar::Mark_NonSpacing category. While the block parser already has a limit (kMaxDiacAfterSymbol = 2) on how many diacritics can follow a base character, the combining marks from Mark_Enclosing (e.g., U+20DD COMBINING ENCLOSING CIRCLE, U+20DE COMBINING ENCLOSING SQUARE, U+20DF COMBINING ENCLOSING DIAMOND, U+20E0 COMBINING ENCLOSING CIRCLE BACKSLASH) and Mark_SpacingCombining categories were not recognized as diacritics and bypassed this limit entirely.

Fix

Extend IsDiacritic() to also recognize QChar::Mark_Enclosing and QChar::Mark_SpacingCombining Unicode categories. This ensures the existing per-character combining mark limit applies to ALL types of Unicode combining characters, preventing the stacking abuse while preserving legitimate use (the limit of 2 combining marks is sufficient for all natural scripts).

Affected file

• ui/text/text.cpp (in desktop-app/lib_ui)

Testing

Text like ы + U+0489 + U+20DD repeated hundreds of times (ы...) should now:

• Display with at most 2 combining marks per base character
• Not cause GPU overload
• Not produce visual corruption

[CLAassistant]

All committers have signed the CLA.

[vserediuk]