[Bug-gnulib] Re: linebreak.c proposed patches for size-calculation overflows

[Simon Josefsson]

Thanks for your opinions.

Paul Eggert <address@hidden> writes:

> Another argument for avoiding _t in user-defined typedefs is that such
> names are reserved by POSIX and or Standard C (I forget which).

This seem like a rather serious argument against '_t'.  Is *_t
reserved, or just some subset?  Now that you mention it, I also
vaguely recall something about _t being reserved for "system types" or
something like that.

I wasn't able to find anything about it in my (possibly early draft
version of) C99 pdf.

>> Another option I have considered is to not use typedef at all, but
>> rather write 'struct foo *foo' instead of 'foo *foo' or 'foo_t *foo'.
>> (I got that idea from GNU lsh.)
>
> The main objection to that is that it requires foo to be a struct
> type; if you later change foo to be some other kind of type you'll be
> stuck.  Other than that it's fine.  In a way it's nicer since struct
> tags are a separate namespace.

Some things, like a library context handle, will never be anything but
a pointer to a struct, so this is probably not a big problem.  It
might even make things more explicit, so the reader understands that
the variable represent complex data.  But I see the problem if you
would over-use 'struct foo *' for every type; some might later change
into 'int*' or something.