syosetu.com

[haporuned]

Character Replacement Issue (�) with Ghostery on Certain Pages

When using Ghostery, some characters on certain pages frequently get replaced by "�" (EFBFBD). For example, the Japanese hiragana character "と" (E381A8) gets replaced by "��" (two "�" characters).

Environment

Browsers: Firefox 133.0.3 (64-bit), Floorp 11.21.0 (64-bit), Waterfox 6.5.2 (64-bit) for Windows, Waterfox 1.0.9.2 for Android
OS: Windows 11, Android 12/11

Steps to Reproduce

The issue sometimes resolves upon reload, but it can persist for the same character.

The affected locations on the page may change.

Once the issue occurs at a certain point, it does not seem to recur further down the page.

The issue is frequently observed on Japanese pages, particularly on "https://syosetu.com/". This site allows users to choose options for text size and line spacing. After the initial render with default settings, the page re-renders with the user's settings. It appears that Ghostery's processing interrupts this re-rendering, causing UTF-8 characters to be split and interpreted as invalid UTF-8 sequences.

[seia-soto]

Hi @haporuned ,

Thank you for reporting the broken page with Ghostery. I just moved this issue to this repository to investigate further and find it in one place. Once I get more details, I'll move this again to proper repository or create another issue in the proper place.

Have a happy year end, and thank you for using Ghostery. Your detailed report always helping us to improve the overall product.

Best

[seia-soto]

Possible breakage with html filtering on Firefox.

[seia-soto]

From the investigation, it seems like this happens when reading Shift JIS encoded characters as UTF8. But it's still curious that how this can happen since syosetu.com is setting its encoding as UTF8.