On Tue, Aug 20, 2002 at 06:02:35PM -0300, Henrique de Moraes Holschuh wrote: > On Tue, 20 Aug 2002, Luca Barbieri wrote: > > Apparently the "different interpretation" is what I was assuming the > > current one. > Yeah, I was in a severe headache for a while because I too knew about the > old interpretation apparently. > > How about the implementing the GNU extension? > It would be useful, yes. But I think it is completely out of the > possibilities for the near future -- you need to get it upstream, and let it > deploy first. > So can we go with versioned symbols (plus -Bsymbolic in key libraries if > just versioned symbols isn't enough for that particular library -- not many > have an API so broken that -Bsymbolic is actually required when versioned > symbols are in use). FWIW, I find that -Bsymbolic tends to be useful in its own right; I've never met anyone who had a good reason for trying to override a library's internal references, but I have seen many cases where not using -Bsymbolic caused namespace conflicts and segfaults. This is a particularly popular source of bugs with Apache/PHP. It's just that -Bsymbolic doesn't solve this particular problem. Steve Langasek postmodern programmer
Attachment:
pgpxcxgeMMy0b.pgp
Description: PGP signature