RE: [Patch] interlibrary dependency parsing in libltdl from CVS HEAD

[Howard Chu]

That's not a problem, since stat() follows symlinks.

  -- Howard Chu
  Chief Architect, Symas Corp.       Director, Highland Sun
  http://www.symas.com               http://highlandsun.com/hyc

> -----Original Message-----
> From: address@hidden [mailto:address@hidden Behalf Of
> Utz-Uwe Haus

> On Sat, Oct 21, 2000 at 03:37:35PM -0500, Albert Chin-A-Young wrote:
> > On Sat, Oct 21, 2000 at 03:03:13PM -0500, Albert Chin-A-Young wrote:
> > > Quick question though. Where dependency_libs is large and references
> > > other .la objects, would it be worthwhile to cache already-dlopen'ed
> > > libraries? If so, what would happen if one of the dlopen'ed libraries
> > > got updated while the program was running? Because we cached the fact
> > > that the library was already loaded, the library would not be reloaded
> > > (though I suppose we could use a stat cache on the modification time
> > > of the file if necessary).
> >
> > Oh, looks like tryall_dlopen already does this.
>
> I haven't looked at the implementation of tryall_dlopen(), but keep in
> mind that one might (ugly solaris scenario follows:) update the
> <lib>.so.version by overwriting while keeping the symbolic link that is
> used to load the library with dlopen(), so a simple stat() may not be
> enough.
>
> Later
> Utz