Re: CVS commit: src/sys/sys

[christos%astron.com@localhost (Christos Zoulas)]

In article <20081231.121418.91326680.he%uninett.no@localhost>,
Havard Eidnes  <he%NetBSD.org@localhost> wrote:
>> On Tue, Dec 30, 2008 at 03:08:09PM +0000, Havard Eidnes wrote:
>>  > Log Message:
>>  > Change va_fsid from long to u_long, so that we can store a dev_t
>>  > there without incurring warnings (which turns into build errors)
>>  > about conversions from unsigned int to long.
>>  > OK'ed by christos@.
>>
>> From the name, I would think it ought to be fsid_t, but it looks as if
>> that's a red herring and it was left partway through a cleanup that
>> intended to change it to fsid_t later. Or something like that.
>
>Hm, yes, possibly.  It appears that fsid_t is a 64-bit non-integral
>type, so changing it to that right now as an alternative to this
>"point fix" would have possibly wide-ranging consequences, and would
>certainly change the binary layout.  On the other hand, I had already
>verified that the simpler fix worked as a "build fix"; I had
>successfully completed several port rebuilds with this fix in place,
>so I concluded that it ought to be good enough for now.

In the christos-time_t branch va_fsid is dev_t and sizeof(dev_t) ==
sizeof(fsid_t) since dev_t is uint64_t. We did not want to change
the binary layout on head that is why we did not widen it.

christos