Re: [PATCH] Define secure_getenv (v2)

["H.J. Lu" <hjl dot tools at gmail dot com>]

On Mon, Jul 30, 2012 at 6:32 AM, Carlos O'Donell
<carlos_odonell@mentor.com> wrote:
> On 7/30/2012 9:07 AM, Florian Weimer wrote:
>> On 07/27/2012 02:58 PM, Carlos O'Donell wrote:
>>> On 7/27/2012 8:53 AM, Florian Weimer wrote:
>>>> On 07/27/2012 02:52 PM, Carlos O'Donell wrote:
>>>>
>>>>> Please tweak the test not to fail and issue a warning when the supplementary
>>>>> group is missing.
>>>>
>>>> Should I print the warning to stderr or stdout?
>>>
>>> Good question. Given that this is a warning about the test itself,
>>> not the thing that the test is testing, I would output this to
>>> stderr. That way a view of stderr would show the warning along with
>>> other warnings about the build.
>>
>> What about this?  Tested on x86_64-redhat-linux-gnu, also with a user without supplementary GIDs.
>>
>
> This looks good to me.
>
> You've converted everything to use stdout, with the exception
> of the detection for a suitable GID, which still prints to
> stderr but because we don't consider it a failure you exit
> with exit(0).
>
> This matches what Siddhesh has summarizing here:
> http://sourceware.org/glibc/wiki/Testing/Testsuite
>

It failed on Linux/x32 since the earliest glibc for x32 is
2.16.  If I understand it correctly, GLIBC_2_16 in

#if SHLIB_COMPAT (libc, GLIBC_2_0, GLIBC_2_16)
compat_symbol (libc, __libc_secure_getenv, __secure_getenv, GLIBC_2_0);
#endif

should be GLIBC_2_17.

-- 
H.J.