Should #ident really be deprecated?

Should #ident really be deprecated?

[Ian Lance Taylor]

Hi Tom, in the fix for PR 22168 you deprecated #ident, in the sense
that gcc now warns about it.  Is that really a good idea?  #ident is a
reasonably widely used extension: codesearch.google.com finds "about"
110,000 uses.  It's supported by other compilers--it was introduced on
System V, I believe, and has been picked up by many other compilers.
It's been in gcc since at least version 1.37.

I don't think there is any reason for us to actually remove #ident, so
I don't see any reason for us to warn that it is deprecated.  We
should certainly warn about it when using -pedantic, but I don't see
any reason to warn about it at any other time.

I reviewed the patch, but I didn't really notice the effect on #ident.
I agree that #assert and #unassert deserve warnings about deprecation.
I'm on the fence about #sccs but I tend to think that we should not
warn about it except with -pedantic.

Ian

Re: Should #ident really be deprecated?

[Tom Tromey]

>>>>> "Ian" == Ian Lance Taylor <iant@google.com> writes:

Ian> Hi Tom, in the fix for PR 22168 you deprecated #ident, in the sense
Ian> that gcc now warns about it.  Is that really a good idea?

No.

I don't remember why I did this, but I assume it was due to the
pre-existing comment above the directive table saying that #ident was
deprecated.

Ian> I don't think there is any reason for us to actually remove #ident, so
Ian> I don't see any reason for us to warn that it is deprecated.  We
Ian> should certainly warn about it when using -pedantic, but I don't see
Ian> any reason to warn about it at any other time.

Yes, I agree.

Ian> I'm on the fence about #sccs but I tend to think that we should not
Ian> warn about it except with -pedantic.

I tend to agree, because my understanding is that #sccs has a long
non-gcc history.

Tom