Minor problems with configure on Solaris 8

[Steve Langasek]

David,

On Fri, Nov 30, 2001 at 12:31:35PM -0500, David Collier-Brown wrote:
>  While looking at the "-h" problem (caused by passing -shared to
> Sun ld), I noticed serveral glitches.

> checking for _read... (cached) yes
> checking for __read... (cached) no

> 	In Unix, this is a bit doubtfull.  The
> 	_ and __ namespaces are reserved for the
> 	implementer's under-the-covers fiddling.
> 
> 	Calling _read instead of read is generally
> 	bad: in a shared library interceptor, one 
> 	normally uses
> 	p = (void *)dlsym((void *)handle, (const char *)"read");
> 
> 	[The Solarii actually use _read for calls to a
> 	read that can't be interecepted...]

It has occurred to me in the past that this would be a way to trump the
glibc maintainers' apparent efforts to thwart library interceptors.  In
addition to needing to detect the right dlopen() calls for the system,
this would also require knowing the location of the libc library on each
platform, correct?  Are there any guarantees that all the necessary
library calls are implemented directly in libc, or could they possibly
be implemented in other libraries that are linked in automatically?

I also suspect that the current scheme works on some platforms where
dlopen() functionality is historically absent or unreliable.

Cheers,
Steve Langasek
postmodern programmer
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 232 bytes
Desc: not available
Url : http://lists.samba.org/archive/samba-technical/attachments/20011130/82ee2eb5/attachment.bin