Messages in this thread | | | Subject | Re: Header files and interfaces | From | (Eric W. Biederman) | Date | 23 Jun 1998 00:24:35 -0500 |
| |
>>>>> "TT" == Theodore Y Ts'o <tytso@MIT.EDU> writes:
TT> There's a difference between the occasional bug (especially in the TT> development kernel) and consciously breaking an old interface, and then TT> when people complain, say that you don't care. The kernel has in TT> general been very careful not to gratutiously break backwards TT> compatibility.
As I understand the story glibc2 has been quite careful as well. llseek was put into some stable release without a header declaration. After realizing he that llseek was not the way to go, Ulrich Drepper decided to never prototype it, to discourage use.
That is fine except that not prototyping things DOES NOT DISCOURAGE USE by autoconf. It instead CREATES BUGS because the return value of llseek without a prototype defaults wrong.
What needs to happen now is that the Ulrich Drepper needs to see his policy failed to DISCOURAGE USE, and CREATED BUGS, and realizing it is a BUG to not prototype functions, prototype it, and find ANOTHER POLICY to discourage use.
Eric
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu
| |