[Webkit-unassigned] [Bug 146473] Errors in read() are not handled in WTF::cryptographicallyRandomValuesFromOS.

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=146473

--- Comment #12 from Keith Miller <keith_miller at apple.com> ---
Comment on attachment 255869
  --> https://bugs.webkit.org/attachment.cgi?id=255869
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=255869&action=review

>> Source/WTF/wtf/OSRandomSource.cpp:67
>> +        if (currentRead < 0 && !(errno == EAGAIN || errno == EINTR))
> 
> I think it would be nicer to check for -1 explicitly, rather than implicitly by including all negatives in the test. The documentation states that read will return -1, 0, or a positive number, and it is nice to be precise.
> 
> You can simplify the logic here a bit and avoid testing currentRead twice like so:
> 
> if (currentRead == -1) {
>     if (!(errno == EAGAIN || errno == EINTR))
>         crashUnableToReadFromURandom();
> } else
>     amountRead += currentRead;
> 
> This helps to clearly separate the failure case from the success case.

That makes sense. Do you think the code you wrote is clearer than:

if (currentRead >= 0)
    amountRead += currentRead;
else if (!(errno == EAGAIN || errno == EINTR))
    crashUnableToReadFromURandom();

I guess it depends on whether or not you think about the success or failure case first. It also, doesn't make the -1 condition explicit.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.webkit.org/pipermail/webkit-unassigned/attachments/20150630/94e15849/attachment.html>