Re: glibc strerrorname_np

[Jonny Grant <jg@jguk.org>]

On 04/11/2021 23:28, Adhemerval Zanella wrote:
> 
> 
> On 04/11/2021 19:52, Jonny Grant wrote:
>> Hi Adhemerval, Thank you for you reply.
>>
>> On 04/11/2021 20:46, Adhemerval Zanella wrote:
>>>
>>>
>>> On 04/11/2021 17:23, Jonny Grant wrote:
>>>> Hi Carlos
>>>> I was pleased to see you added strerrorname_np()
>>>>
>>>> May I ask, I couldn't find the file your implementation is in - could you point it out to me in the glibc repository please?
>>>
>>> It is on string/strerrorname_np.c, which calls __get_errname() defined at
>>> stdio-common/errlist.c.
>>
>> Ok yes, I see in glibc/stdio-common/errlist.c the __get_errname() implementation.
>>
>>>> I noticed on the man page it may return NULL, which is a shame, as then it means we always need to check that before using in every printf etc :-
>>>>
>>>> printf("err %s\n", strerrorname_np(myerr)?strerrorname_np(myerr), "Unknown err");
>>>
>>> I didn't considered printf() when I added strerrorname_np(). Maybe an empty string ("")
>>> would be better than NULL.
>>
>> Yes, "" empty string sounds great.
>>  
>>>> I'd done my own version a while ago as strerrno_s(), and assumed I could never get it accepted anywhere like glibc.
>>>> Probably I should have tried to submit it to glibc!
>>>> https://github.com/jonnygrant/safec/blob/master/strerrno.c
>>>>
>>>> Would something like my implementation ever be accepted?
>>>> errno_t strerrno_s(char * const buf, const rsize_t buflen, const errno_t errnum)
>>>
>>> So this is basically:
>>>
>>>   int strerrno_s (char *buf, size_t buflen, int errnum)
>>>   {
>>>     const char *r = strerrorname_np (errnum);
>>>     __snprintf (buf, buflen, "%s", r == NULL ? "" : r);
>>>     return errnum;
>>>   }
>>>
>>> I don't see much gain on adding another wrapper to format errno, the idea of
>>> strerrorname_np() was to provide a async-signal-safe way to map errno to
>>> string (by avoiding translation).
>>
>> Yes, I should migrate to your strerrorname_np() version soon. If you could change it to return "" instead of NULL, that would be very much appreciated to avoid any accidental SEGV.
> 
> This is most likely an ABI break and I really like to avoid it since it would
> require an new version and another implementation for such symbol.
> 

Hi Adhemerval

Thank you for your reply. Personally I understood an ABI break would be the return type, the name, or the parameters. But the proposed change is not so. Changing to return a string, should be fine.

ie, in relation to strerror() C99 and POSIX.1-2008 require the return value to be non-NULL. (my view is it is always better not to return a NULL from such string functions that could then cause a SEGV.

strerror(1000) returns a string "Unknown error 1000"

Better to simply align with glibc strerror() approach?

Feels like there is still time to change it, as it is _np. Aligning with strerror(), or just "" as you had mentioned seems reasonable.


https://man7.org/linux/man-pages/man3/strerror.3.html

It's common for some returns to change, eg glibc 2.13 changed strerror_r() behaviour to return the actual error code, as opposed to returning -1 and setting errno.

Cheers
Jonny