Fw: Issue: headers may use non-reserved identifiers
Richard Damon
Richard@Damon-Family.org
Thu Apr 21 02:48:24 GMT 2022
On 4/20/22 6:11 PM, C Howland wrote:
>> ------------------------------
>> *From:* Newlib on behalf of Pavel M <pavel.morozkin@gmail.com>
>> *Sent:* Wednesday, April 20, 2022 4:35 PM
>> *To:* newlib@sourceware.org <newlib@sourceware.org>
>> *Subject:* Issue: headers may use non-reserved identifiers
>>
>>
>> Hi all,
>>
>> Issue: headers may use non-reserved identifiers.
>>
>> Example:
>> #define _reent 0
>> #include <stdio.h>
>>
>> $ gcc t567.c -std=c11
>> t567.c:1:16: error: expected ‘{’ before numeric constant
>> 1 | #define _reent 0
>> | ^
>> t567.c:1:16: error: expected ‘{’ before numeric constant
>> 1 | #define _reent 0
>> | ^
>> and so on...
>>
>> Per C11 _reent, _on_exit_args, etc. are non-reserved identifiers.
>>
>> Consider fixing.
>>
>> P.S. Good if it leads to compile time errors, not good if it doesn't (wrong
>> translation unit produced => wrong code generated => wrong runtime
>> behavior).
>>
>> --
>> Pavel
>> ------------------------------
>>
>
> No, the headers do not use non-reserved identifiers. Look at the standard
> again, as the second dash item in section 7.1.3 Reserved Identifiers
> reserves all identifiers that start with underscore. The example program
> violates the standard. In short, user programs may not use identifiers
> that start with underscore. (More precisely there are some very-limited
> cases in which they can, but preprocessor identifiers is not one of them.)
Slight correction, identifies that begin with an underscore and followed
by a lowercase letter or a number are reserved only at 'file scope'
level, so user programs CAN use them as local variable names inside some
scope (at least those that don't match the previous cluase of an
underscore followed by an uppercase letter or another underscore as
those are reserved in all conditions).
Now, in the example, the user program used it for a preprocessor symbol
which conflicts with the reservation for that identifier for the
implementations use.
--
Richard Damon
More information about the Newlib
mailing list