[openib-general] mvapich2 ofed 1.2 problem

[Shaun Rowland]

Doug Ledford wrote:

> It didn't loose the information, it never had it.  When you link both
> libs against the application binary, the linker is resolving linkups and
> writing that into the resulting application binary output, but unless
> it's allowed to write into the libsrqtest.so binary and modify *it's*
> link table, that particular versioning information can't be written.

I thought that this might be the case, but I had never run into this
before. Thanks for clearing that up.

> Obviously, if every gcc compile that touched a shared library as a
> source object file also attempted to write back to that source object
> file, people would be very surprised when their attempt to link an
> application failed due to permission problems on the shared library.

Yes. I thought perhaps it would use the default ABI when the symbols
were resolved when making the binary, but as I said, I've never seen
this issue before. Clearly, it is working as you've described, and that
one thought I had seems not to make sense. Even when I tried making my
shared library the way I thought it should be done the first time, I
linked libibverbs into it at shared library creation time. Only when I
saw the difference did I try waiting until building the application.

> Yes.  If you want symbol information in a shared lib that uses other
> shared libs, then they have to be linked at .so creation time, not at
> application creation time.

I can make this happen.  I am testing it now.
-- 
Shaun Rowland	rowland at cse.ohio-state.edu
http://www.cse.ohio-state.edu/~rowland/