Joerg Schilling wrote: > let us look at the standard.... > > The file argument is used to construct a pathname to the object file. > If > file contains a slash character, the file argument is used as the > pathname > for the file. Otherwise, file is used in an implementation-defined > manner > to yield a pathname. > > See http://www.opengroup.org/onlinepubs/009695399/functions/dlopen.html
So you're saying that if there is a "/" anywhere in the dlopen() string, then the *whole* string would be taken as is. That's also an option. Although I've long suggested that folks try to avoid full pathnames. Simple filenames allow the use of RPATH ($(ORIGIN)), overloading by LD_LIBRARY_PATH for experimentation, and even crle(1). You loose a great deal of flexibility (even if it's just for your debugging environment) when you use full pathnames. Presently we would split the string into ":" components, and apply the "/" rule to any individual component. I was really hoping to hear a little more from the Wine folks (Robert?) as to which of our suggestions would be beneficial to them. I don't want to invent something that doesn't help solve the problem at hand. -- Rod
