Bug 128068 – graphics/StringTruncator.cpp:172: possible bad array index ?

• Component: Platform
• Status: RESOLVED INVALID

[David Binderman]

I just ran the static analyser "cppcheck" over the source
code of webkitgtk-2.3.4

It said many things, including

[Source/WebCore/platform/graphics/StringTruncator.cpp:172]: (style) Array index 'adjustedStartIndex' is used before limits check.

Source code is

    // Strip single character after ellipsis character, when that character is preceded by a space
    if (adjustedStartIndex < length && string[adjustedStartIndex] != space
        && adjustedStartIndex < length - 1 && string[adjustedStartIndex + 1] == space)
        ++adjustedStartIndex;

Maybe

    if (adjustedStartIndex < length && string[adjustedStartIndex] != space
        && adjustedStartIndex > 0 && string[adjustedStartIndex - 1] == space)

would be better code

[Daniel Bates]

(In reply to comment #0)
> I just ran the static analyser "cppcheck" over the source
> code of webkitgtk-2.3.4
> 
> It said many things, including
> 
> [Source/WebCore/platform/graphics/StringTruncator.cpp:172]: (style) Array index 'adjustedStartIndex' is used before limits check.
> 
> Source code is
> 
>     // Strip single character after ellipsis character, when that character is preceded by a space
>     if (adjustedStartIndex < length && string[adjustedStartIndex] != space
>         && adjustedStartIndex < length - 1 && string[adjustedStartIndex + 1] == space)
>         ++adjustedStartIndex;
> 
> Maybe
> 
>     if (adjustedStartIndex < length && string[adjustedStartIndex] != space
>         && adjustedStartIndex > 0 && string[adjustedStartIndex - 1] == space)
> 
> would be better code

I'm unclear how your proposed change would work given the omission of the body of the if-statement.

[Ahmad Saleem]

We still have this code:

https://searchfox.org/wubkat/source/Source/WebCore/platform/graphics/StringTruncator.cpp#178

Do we need to action anything?

[Alexey Proskuryakov]

I don't think that any action is needed. Maybe this code can be rewritten to be slightly nicer, but it's reasonable to only check for buffer overrun here, not for underrun.