Hi, On Mon, Mar 31, 2014 at 12:27 PM, Chris Barker <chris.bar...@noaa.gov> wrote: > On Mon, Mar 31, 2014 at 12:05 PM, Matthew Brett <matthew.br...@gmail.com> > wrote: >> >> > that part, yes, but isn't Linux too much of a varying target for there >> > to be >> > any point anyway? >> >> You mean, the /usr/lib stuff varies too much, so that any copied >> dynamic libraries would have little chance of binary compatibility >> with the system libs? > > > exactly. > >> > The "usual" way is to put the dll next to where it is needed. I _think_ >> > a >> > when a one dll (the pyton extension) is linked to another one, the first >> > place windows looks is right next to the one loading it -- same as for >> > dlls >> > linked to main executables. >> >> I had assumed from [1] is that it's the path of the executable not the >> loading DLL that is on the DLL search path, but I might well be wrong > > > I could be wring, too -- I'm pretty sure I tested this at some point, but It > could be getting lost in the fog of memory.
I am hopelessly lost here, but it looks as though Python extension modules get loaded via hDLL = LoadLibraryEx(pathname, NULL, LOAD_WITH_ALTERED_SEARCH_PATH); See: http://hg.python.org/cpython/file/3a1db0d2747e/Python/dynload_win.c#l195 I think this means that the first directory on the search path is indeed the path containing the extension module: http://msdn.microsoft.com/en-us/library/windows/desktop/ms682586(v=vs.85).aspx#alternate_search_order_for_desktop_applications So I'm guessing that it would not work putting DLLs into the 'DLLs' directory - unless the extension modules went in there too. I _think_ (David ?) this means it would not work to copy the dependent DLLs into sys.exec_prefix Cheers, Matthew _______________________________________________ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion