On Feb 24, 2015, at 4:45 PM, Paul Hargrove <phhargr...@lbl.gov> wrote: > > Check for dlfcn.h and the dlopen symbol in -ldl. > > Then the paranoid part of me wants to note that since you don't try using > dlopen() in the configure tests you risk encountering platforms with > non-functional/non-conforming implementations. For instance stubs that > return w/ errno=ENOSYS or which don't support some flags you want/need to > pass. I suppose one can safely ignore that possibility unless/until one > encounters such a beast.
I thought about this: but do such systems really exist any more? > On a related note: if dlopen() does fail at runtime, will things just > "magically" fallback to libltdl (assuming both were found at configure time). No; I currently have it only build one of the two components. I'd have to think about it a bit, but it might be possible to build both, and if one fails to perform an action (e.g., open a file), try the other. Hmm.... > The Solaris-10/SPARC builds will take a couple hours, but I should have > results well before I quit for the day. So, you should certainly have email > about any "kinks" I find before you start any merge on Wed. Great. I love getting Hargroved. -- Jeff Squyres jsquy...@cisco.com For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/
