Discussion:
opensm copy of ib_types.h
(too old to reply)
Sean Hefty
2009-02-19 20:09:26 UTC
Permalink
Does anyone know why the opensm copy of ib_types.h has:

#ifndef WIN32

to basically exclude the entire file?
Smith, Stan
2009-02-19 20:37:15 UTC
Permalink
Post by Sean Hefty
#ifndef WIN32
to basically exclude the entire file?
_______________________________________________
ofw mailing list
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ofw
Another question - Why do we have a duplicate file in the 1st place?

Opensm version of ib_types.h vs. trunk\inc\iba\ib_types.h

Possibly a way to support 'newer' openSM defs without updating inc\iba\ib_types.h ?
Weak...
Sean Hefty
2009-02-19 20:53:11 UTC
Permalink
Post by Smith, Stan
Opensm version of ib_types.h vs. trunk\inc\iba\ib_types.h
I'm going down the path of updating inc\iba\ib_types.h to contain the
definitions that I need (for the diags). This will change some structure
definitions, and I'll update anything in the tree that uses them.

- Sean
Smith, Stan
2009-02-19 22:24:56 UTC
Permalink
Post by Sean Hefty
Post by Smith, Stan
Opensm version of ib_types.h vs. trunk\inc\iba\ib_types.h
I'm going down the path of updating inc\iba\ib_types.h to contain the
definitions that I need (for the diags). This will change some
structure definitions, and I'll update anything in the tree that uses
them.
- Sean
Good news!
The implication being the OpenSM version of ib_types.h can eventually be removed?
Smith, Stan
2009-02-19 22:27:07 UTC
Permalink
Post by Smith, Stan
Post by Sean Hefty
Post by Smith, Stan
Opensm version of ib_types.h vs. trunk\inc\iba\ib_types.h
I'm going down the path of updating inc\iba\ib_types.h to contain the
definitions that I need (for the diags). This will change some
structure definitions, and I'll update anything in the tree that
uses them.
- Sean
Good news!
The implication being the OpenSM version of ib_types.h can eventually be removed?
Could OpenSM's ib_types_extended.h also be purged?
Post by Smith, Stan
_______________________________________________
ofw mailing list
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ofw
Hal Rosenstock
2009-02-19 22:33:06 UTC
Permalink
Post by Smith, Stan
Post by Sean Hefty
Post by Smith, Stan
Opensm version of ib_types.h vs. trunk\inc\iba\ib_types.h
I'm going down the path of updating inc\iba\ib_types.h to contain the
definitions that I need (for the diags). This will change some
structure definitions, and I'll update anything in the tree that uses
them.
- Sean
Good news!
The implication being the OpenSM version of ib_types.h can eventually be removed?
Doesn't OpenSM need that version of ib_types.h (and it does use a lot
more than the diags do) ?

I do agree with the goal of getting this down to one copy of this file.

-- Hal
Post by Smith, Stan
_______________________________________________
ofw mailing list
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ofw
Sean Hefty
2009-02-19 22:37:03 UTC
Permalink
Post by Hal Rosenstock
Doesn't OpenSM need that version of ib_types.h (and it does use a lot
more than the diags do) ?
Simply put, I need to update one of the two files. I'm choosing
inc\iba\ib_types.h because of the weirdness of the opensm files (#ifndef WIN32
and ib_types_extended.h).

- Sean
Sean Hefty
2009-02-19 22:33:51 UTC
Permalink
Post by Smith, Stan
The implication being the OpenSM version of ib_types.h can eventually be removed?
Eventually, but I'm not trying to update the two versions completely. I'm only
updating the structures that cause issues building the diags, in particular
ibsendtrap and saquery.

The structure definitions in the winof ib_types.h files differ from those used
by OFED, which includes spec updates. Unfortunately, the path record structure
is one of the structures that needs updating, so I'll see how that goes.

I guess once I get saquery operational, I can see what sort of chaos results
from opensm including the other ib_types.h and if the problems can be fixed
within my lifetime.

- Sean
Smith, Stan
2009-02-19 22:37:44 UTC
Permalink
Post by Sean Hefty
Post by Smith, Stan
The implication being the OpenSM version of ib_types.h can
eventually be removed?
Eventually, but I'm not trying to update the two versions completely.
I'm only updating the structures that cause issues building the
diags, in particular ibsendtrap and saquery.
The structure definitions in the winof ib_types.h files differ from
those used by OFED, which includes spec updates. Unfortunately, the
path record structure is one of the structures that needs updating,
so I'll see how that goes.
FYI - In the Windows opensm version of ib_types.h, I modified the path record fields to what saquery was expecting. Ensuing mods were applied to Windows OpenSM code base.
Post by Sean Hefty
I guess once I get saquery operational, I can see what sort of chaos
results from opensm including the other ib_types.h and if the
problems can be fixed within my lifetime.
- Sean
Hefty, Sean
2009-02-19 22:40:46 UTC
Permalink
Post by Smith, Stan
FYI - In the Windows opensm version of ib_types.h, I modified the path record
fields to what saquery was expecting. Ensuing mods were applied to Windows
OpenSM code base.
I started down this path, but hit into the issue that IB_SUCCESS, among other things, was not defined. How did you work around this issue?
Smith, Stan
2009-02-19 23:54:08 UTC
Permalink
Post by Hefty, Sean
Post by Smith, Stan
FYI - In the Windows opensm version of ib_types.h, I modified the
path record fields to what saquery was expecting. Ensuing mods were
applied to Windows OpenSM code base.
I started down this path, but hit into the issue that IB_SUCCESS,
among other things, was not defined. How did you work around this
issue?
Svn.1925 ulp\opensm\user\include\iba\ib_types.h is when the I put in the saquery mods.

I don't recollect having a problem with IB_SUCCESS. Although I do remember an include ordering issue such that the openSM version of ib_types.h needed to come before inc\iba\ib_types.h.
Also I was working from saquery source before Shasha purged the osm_vendor routines from saquery.

If you recreate the svn.1925 opensm\user\include\iba\ib_types.h it may apply to inc\iba\ib_types.h and save you some work.
Continue reading on narkive:
Loading...