[libcxx] r174642 - Belt and suspenders when calling sysconf
zarzycki at apple.com
Thu Feb 7 11:10:36 PST 2013
To be clear, I am *not* objecting to this patch. It is clearly safe in this specific case.
My point is that replacing "== -1" with "< 0" is a common POSIX anti-pattern. If negative values do not make sense, then arguably an assert() is appropriate if one wants to be paranoid.
On Feb 7, 2013, at 2:04 PM, Howard Hinnant <hhinnant at apple.com> wrote:
> Fwiw, Marshall discussed this change with me before committing. That being said, I have no objection to reverting it either. And either way we go, this code does not execute on Apple platforms (#ifdef'd out).
> On Feb 7, 2013, at 1:54 PM, Dave Zarzycki <zarzycki at apple.com> wrote:
>> FYI — While safe in this specific example, this is not a correct transform to make in the general case. POSIX very clearly specifies that -1 is the error code for this and similar APIs. This matters for APIs like mmap() which can return "negative" values that not an error.
>> On Feb 7, 2013, at 1:48 PM, Marshall Clow <mclow at qualcomm.com> wrote:
>>> Author: marshall
>>> Date: Thu Feb 7 12:48:09 2013
>>> New Revision: 174642
>>> URL: http://llvm.org/viewvc/llvm-project?rev=174642&view=rev
>>> Belt and suspenders when calling sysconf
>>> Modified: libcxx/trunk/src/thread.cpp
>>> URL: http://llvm.org/viewvc/llvm-project/libcxx/trunk/src/thread.cpp?rev=174642&r1=174641&r2=174642&view=diff
>>> --- libcxx/trunk/src/thread.cpp (original)
>>> +++ libcxx/trunk/src/thread.cpp Thu Feb 7 12:48:09 2013
>>> @@ -69,7 +69,9 @@ thread::hardware_concurrency() _NOEXCEPT
>>> long result = sysconf(_SC_NPROCESSORS_ONLN);
>>> // sysconf returns -1 if the name is invalid, the option does not exist or
>>> // does not have a definite limit.
>>> - if (result == -1)
>>> + // if sysconf returns some other negative number, we have no idea
>>> + // what is going on. Default to something safe.
>>> + if (result < 0)
>>> return 0;
>>> return static_cast<unsigned>(result);
>>> #else // defined(CTL_HW) && defined(HW_NCPU)
>>> cfe-commits mailing list
>>> cfe-commits at cs.uiuc.edu
>> cfe-commits mailing list
>> cfe-commits at cs.uiuc.edu
More information about the cfe-commits