Messages in this thread |  | | | Date | Tue, 7 Dec 1999 23:57:53 -0800 (PST) | | From | Linus Torvalds <> | | Subject | Re: What I suspect |
| |
On 7 Dec 1999, Ulrich Drepper wrote: > Linus Torvalds <torvalds@transmeta.com> writes: > > > But is that really even an issue? > > People are even complaining when I fix bugs in the behaviour of the > ELF loader.
Hey, people always complain when something changes. No way to get around that..
> I don't even want to start thinking about the reaction > when suddenly well documented and correctly implemented functionality > is changing. Don't even think about going this road.
You're implying that my suggestion would not be "well documented or corectly implemented". I don't agree. If you're adding _new_ weak functions to a standard library, I don't think you want old binaries to suddenly change behaviour just because they happened to use a symbol that now clashes with the new library (but didn't even use to exist).
Standard libraries shouldn't add new symbols anyway in a perfect world, but if they do, you _definitely_ don't want to change the way an old binary works. I can see nothing but chaos from that.
So I still think pre-linking should be perfectly fine =without= checking weak symbols like Davem suggested. In fact, I'd argue that re-checking is the wrong thing to do from a "least surprises" standpoint.
(Of course, if the link changes dynamically due to miscompare of the pre-linked address or because new prelinked libraries are added, then the full linking has to be done - it just shouldn't be necessary for the common case).
Linus
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
|  |