Keith M Wesolowski wrote:
> On Mon, Jul 21, 2008 at 09:44:50PM +0300, Cyril Plisko wrote:
>
>
>> I have a real case here with customer' C++ application that happens to
>> be related to SCSI. They are evaluating support for Solaris
>> Express/OpenSolaris. The offending header file is
>> /usr/include/sys/scsi/targets/stdef.h [1]. The structure member
>> "explicit" breaks compilation with C++. Apparently this is newish
>> code, that was added in rev 5628 [2], so they never saw this with
>> Solaris 10.
>> So does this warrants filing a bug against stdef.h ?
>>
>
> Yes. The lack of prefixing in most of the SCSI subsystem is nasty
> anyway; the fact that it breaks real-world code is just a good excuse
> to fix some of it sooner rather than later.
>
> The only thing I would be cautious of is that there are no consumers
> other than st.c. I couldn't find any using cscope but I also couldn't
> find an ARC case that covered the putback that introduced these
> structures. I couldn't find any references to these structures in man
> pages so presumably they're private but it would be good to ask the
> engineer to be sure.
>
>
One of the things that I think we (Sun) do poorly is expose header files
(and structures) that have no business being in /usr/include. For
example, device-driver private headers that only are used by the driver
itself have no business being exported in /usr/include. (Most NIC
drivers, for example, fall into this category. I suspect this is true
for many other target/leaf drivers as well.)
-- Garrett
_______________________________________________
opensolaris-code mailing list
opensolaris-code@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
