> #include <string>
> #include <dlfcn.h>
>
> const std::string& GetPath();
>
> namespace {
> std::string reallyGetPath() {
> Dl_info dl_info;
> dladdr((void*)GetPath, &dl_info);
> return dl_info.dli_fname;
> }
> }
>
> const std::string& GetPath() {
> static const std::string myPath = reallyGetPath();
> return myPath;
> }
>
> The compiler adds locking to the initialization of myPath.
Everyone, thank you for the replies and help. Jonathon, I did, eventually, take the advice given. I really didn't want to lose the "discovery of path on initialization of the library" and that's why I posted as I did last week. I realized, however, that all of this was very much in vain. Also, I have been reeducated in something which I had completely forgotten. That is, I'd forgotten that the compiler initializes static function variables only once as is depicted above. The solution I wrote has an extraneous conditional in it (the very thing I wanted most to avoid).
I have one question further. I tried using this approach because of what I read in the GCC manuals, "Functions with these attributes are useful for initializing data that is used implicitly during the execution of the program. (taken from http://gcc.gnu.org/onlinedocs/gcc/Function-Attributes.html)" I wish only to understand more fully, if initialization of data, as in this case, is so problematic, why is it such a good thing? Is it because my data variable wasn't truly global, but instead translation-unit-only? What am I missing from what the manual is describing?
Thanks for all the help.
Andy
