hmm ... RT has a weird bug that added an empty comment whilst changing 
ticket status. Please ignore it. 
 
[EMAIL PROTECTED] - Wed Sep 24 07:48:10 2003]: 
> Anyway, here's a patch for MacOS X, the patch for dso_dlfcn.c could be 
> implemented in other ways also. There would be no need for ifdef's if 
> dlfcn_bind_var and dlfcn_bind_func would just probe for underscored 
>    symbol 
> names at first, and then without it, if the first probe fails. I'd 
>    rather 
> see the later method used, if possible. 
 
I don't like the idea of generic dlfcn handlers magically checking for translated 
versions of symbols before looking for the "as-is" names they were passed. In 
particular, this gets grimy if a loaded library defines symbols that differ only by 
how many leading underscores they have! 
 
IMHO: there are two acceptable solutions - (i) the one you've sketeched out, 
which is to test with a prefixed underscore *if* the configuration indicates we 
are building for a platform that does this (in which case the secondary test 
without the underscore is presumably useless - and worth using an "#else" 
clause to eliminate it perhaps??), or (ii) implement a distinct DSO_METHOD 
that piggy-backs DSO_METHOD_dlfcn() by performing the underscore 
prefixing before handing control over. 
 
For now, I think the -DDLFCN_NEED_UNDERSCORE approach is fine, and 
(ii) would be total overkill anyway. Could you please regenerate your diff 
against a recent 0.9.7-stable snapshot (from ftp://ftp.openssl.org/snapshot/) 
and test that this still has the desired effect on MacOS-X? The diff in this 
ticket didn't apply cleanly to my 0.9.7-checkout, and I have no MacOS-X box to 
confirm with if I meddle with the patch manually. Also, please take a look at 
the snapshot for the development head too - we will have to commit an 
equivalent there so the MacOS-X support doesn't go missing in the next major 
release. The underscore change to dso_dlfcn.c will port trivially enough, but the 
Configure and Makefile.org changes won't - the development head uses a 
different scheme for building shared libraries. 
 
Cheers, 
Geoff 
 
--  
Geoff Thorpe, RT/openssl.org 
______________________________________________________________________
OpenSSL Project                                 http://www.openssl.org
Development Mailing List                       [EMAIL PROTECTED]
Automated List Manager                           [EMAIL PROTECTED]

Reply via email to