[PATCH] [libc++] Don't return uninitialized data from random_device::operator()

[David Majnemer]

Updated to address review comments.


On Wed, May 28, 2014 at 7:23 AM, Marshall Clow <mclow.lists at gmail.com>wrote:

>
> On May 27, 2014, at 11:44 AM, David Majnemer <david.majnemer at gmail.com>
> wrote:
>
> On Tue, May 27, 2014 at 7:45 AM, Marshall Clow <mclow.lists at gmail.com>
>>  wrote:
>>
>>>
>>> On May 20, 2014, at 2:25 AM, David Majnemer <david.majnemer at gmail.com>
>>> wrote:
>>>
>>> Oops, sent out the wrong version of this patch.  Attached is what I
>>> intended to send.
>>>
>>>
>>> On Tue, May 20, 2014 at 12:28 AM, David Majnemer <
>>> david.majnemer at gmail.com> wrote:
>>>
>>>> random_device::operator() as currently implemented does not correctly
>>>> handle errors returned by read.  This can result in it returning
>>>> uninitialized data.
>>>>
>>>> To fix this, wrap the call to read in a loop.
>>>>
>>>
>>> I like this; but can you think of any way to test it?
>>>
>>
> I've added a test for the EOF case.
>
> I have written a test for EINTR case but I did not included it because
> it's inherently not reliable. It checks the output of operator() to see if
> signals resulted in us getting a lot of zero return results.
>
>
> I get a couple of signedness warnings when building:
>
> + for FILE in '../src/*.cpp'
> + /Sources/LLVM/bin/bin/clang++ -c -g -Os -arch i386 -arch x86_64
> -nostdinc++ -std=c++11 -fstrict-aliasing -Wall -Wextra -Wshadow
> -Wconversion -Wpadded -Wstrict-aliasing=2 -Wstrict-overflow=4 -I../include
> ../src/random.cpp
> ../src/random.cpp:78:14: warning: implicit conversion changes signedness:
> 'ssize_t' (aka 'long') to 'unsigned long' [-Wsign-conversion]
>         i += s;
>           ~~ ^
> 1 warning generated.
> ../src/random.cpp:78:14: warning: implicit conversion changes signedness:
> 'ssize_t' (aka 'long') to 'unsigned long' [-Wsign-conversion]
>         i += s;
>           ~~ ^
> 1 warning generated.
>
> I get that read can only return -1, 0, and positive numbers, and you’re
> already checking for -1 and 0, so I think that you can safely cast s to a
> size_t before doing the addition.
>
>
> Also, from a readability standpoint, wouldn’t a while loop be better than
> the for loop?
> Something like this:
>
>     while (n > 0)
>     {
>         ssize_t s = read(__f_, p, n);
>         if (s == 0)
>             __throw_system_error(ENODATA, "random_device got EOF");
>         if (s == -1)
>         {
>             if (errno != EINTR)
>                 __throw_system_error(errno, "random_device got an
> unexpected error");
>             continue;
>         }
>         i -= (size_t) s;
>         p += (size_t) s;
>     }
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20140528/d07ebf26/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: t.patch
Type: application/octet-stream
Size: 1524 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20140528/d07ebf26/attachment.obj>