[EMAIL PROTECTED] wrote on Fri, 01 Sep 2006 09:56 -0500:
> I ran into some compiler warnings on the openib code that I hadnt
> previously seen before after checking out the CVS head.
>
> src/io/bmi/bmi_ib/openib.c: In function openib_post_sr_ack:
> src/io/bmi/bmi_ib/openib.c:545: warning: missing initializer
> src/io/bmi/bmi_ib/openib.c:545: warning: (near initialization for sr.wr)
> ---
> I'm not sure how to fix this one immediately because we do not use the
> sr.wr union
> when posting ack messages.
No clue why the compiler would complain about this. Since we're
using named initializers, and don't touch sr.wr or anything inside
it, we don't have to initialize everything. Hopefully it's a
passing compiler bug. My gcc-4.0.2 on x86-64 doesn't complain
nor do any of the older ones I've been using.
> src/io/bmi/bmi_ib/openib.c:836: warning: enumeration value
> âIBV_EVENT_CLIENT_REREGISTERâ not handled in switch
> --- I 've attached a patch to fix this warning, but didnt go through and
> check the vapi-IB equivalent, we may want to look into that at some
> point, though not sure if its necessary.
Yeah, this has appeared in recent libibverbs, but is absent in 1.0
that many have installed (and that ships with Fedora, e.g.). I
applied your patch but also along with the configure-time
infrastructure. It's not as easy as just adding an #ifdef since
the constant is defined through an enum. Someday I'll figure out
what this event means and add code to deal with the situation.
-- Pete
_______________________________________________
Pvfs2-developers mailing list
Pvfs2-developers@beowulf-underground.org
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-developers
