[illumos-Developer] Ugh... xdr_{float,double} problems

Jason King jason.brian.king at gmail.com
Sun Jan 30 14:38:15 PST 2011 

On Sun, Jan 30, 2011 at 4:16 PM, Garrett D'Amore <garrett at nexenta.com> wrote:
> On Sun, 2011-01-30 at 16:07 -0600, Jason King wrote:
>> One thing I did notice is that they are typically surrounded by
>> #ifdef __PRAGMA_REDEFINE_EXTNAME
>> #endif
>> blocks
>>
>> I will update the header file to match the existing convention.
>>
>> One other thing to note -- it appears the existing xdr.h doesn't pass
>> the cstyle checks (independent of my changes).
>>
>> Some of them I can clean up as it's just some spacing issues, but it
>> complains a lot about using 'u_long' vs. 'ulong_t' and similar such
>> differences with a few other integral types.  I'm more hesitant to
>> change those without some additional guidance.  I don't think they
>> would break anything by changing them.  ISTR that C isn't _that_
>> strict about typing (so using ulong_t vs u_long shouldn't make a
>> difference), but it's one of those hairy areas that I'm not sure
>> about.
>
> You are not responsible for fixing pre-existing issues.  Your new code
> should be clean, and if its easy to clean stuff, then you should.  But
> not if its going to be really hairy and risky.
>
> That said, ulong == u_long == ulong_t where C is concerned.  These are
> all typedef'd to "unsigned long", and you can change them as you need to
> to pass C style without any risk.

It's small enough in number that I threw it in.

http://cr.illumos.org/view/ka87apne/ has all the updates (sans hg recommit).

I've just been doing a dmake install/dmake lint with the current fixes
(which all come back ok), so now with all the issues fixed, I've
kicked off a full nightly on x86 & sparc to be 100% sure everything's
good.

>
>>
>> If not, I could add the /* {BEGIN,END} CSTYLED */ tags as an
>> alternative (or just ignore it altogether if everyone's happy with
>> that as well).
>
> No, we'll just waive the cstyle output in this case.  Someday someone
> *should* go back and fix that.
>
>        - Garrett
>
>>
>> On Sun, Jan 30, 2011 at 3:38 PM, Garrett D'Amore <garrett at nexenta.com> wrote:
>> > If that particular pragma is so widely used, then lets just go ahead and
>> > keep it for now.
>> >
>> >        - Garrett
>> >
>> > On Sun, 2011-01-30 at 15:33 -0600, Jason King wrote:
>> >> On Sun, Jan 30, 2011 at 2:52 PM, Garrett D'Amore <garrett at nexenta.com> wrote:
>> >> > General review feedback:
>> >> >
>> >> > No spacing between # and CPP directives, please.
>> >>
>> >> Fixed.
>> >>
>> >> >
>> >> > I'm a bit worried about pragma redefine_extname -- what happens to that
>> >> > on gcc or Studio?  Do they both support it?
>> >>
>> >> They should -- it's used extensively throughout the code already
>> >> (stdio, sockets, etc. about 150 times), studio definitely does, if gcc
>> >> doesn't there would be much bigger problems :)
>> >>
>> >> >
>> >> > I'm thinking #error might be better here.... it would force people
>> >> > working on alternate processor ports to adapt the code...
>> >>
>> >> In that case, should we just ditch the portable implementation altogether?
>> >>
>> >> >
>> >> >        - Garrett
>> >> >
>> >> > On Sun, 2011-01-30 at 14:27 -0600, Jason King wrote:
>> >> >> Yeah, basically I'm a dumbass and shouldn't try to bugfix during weeks
>> >> >> where I'm sleep deprived (my apologies to everyone for that).  Dr.
>> >> >> Grabowski was right
>> >> >>
>> >> >> I see the problem, and was able to get access to a S10 (sparc & x86)
>> >> >> box to obtain a known good XDR bytestream (of floats and doubles) to
>> >> >> test against so the test can be more definitive.
>> >> >>
>> >> >> I have a webrev of the fix here: http://cr.illumos.org/view/kc6gu36g/
>> >> >> The #pragma redefine_extname bit was a suggestion from Gordon (just to
>> >> >> help people porting to other platforms out).
>> >> >>
>> >> >> I really need to run errands now but will be back later to post the
>> >> >> test code (short version, I take the known good stream from s10
>> >> >> x86/sparc, decode it, compare values, then encode the values to
>> >> >> another stream and compare vs. trying to 'guess' the correct values in
>> >> >> the previous test), and take care of the rest.
>> >> >>
>> >> >> On Sat, Jan 29, 2011 at 10:05 PM, Garrett D'Amore <garrett at nexenta.com> wrote:
>> >> >> > Any update on this?  I've not had a chance yet to backout your change..
>> >> >> > I'm not sure what the right thing to do here is... I'd rather have a
>> >> >> > "real" fix.
>> >> >> >
>> >> >> >        - Garrett
>> >> >> >
>> >> >> > On Fri, 2011-01-28 at 12:10 -0600, Jason King wrote:
>> >> >> >> Ok, however things will still be broken, my fix just breaks (or more
>> >> >> >> precisely, breaks everyone's workarounds) faster.  I can probably have
>> >> >> >> a true fix later tonight as well.
>> >> >> >>
>> >> >> >> On Fri, Jan 28, 2011 at 10:58 AM, Garrett D'Amore <garrett at nexenta.com> wrote:
>> >> >> >> > Ok, I'll backout the changes.  Can you please file a bug?  I probably won't
>> >> >> >> > get around to backing this change out until tonight as I'm traveling and on
>> >> >> >> > limited battery power.
>> >> >> >> >
>> >> >> >> >    - Garrett
>> >> >> >> >
>> >> >> >> > On 01/28/11 08:48 AM, Jason King wrote:
>> >> >> >> >>
>> >> >> >> >> After lengthy discussions, and even more digging... it looks like
>> >> >> >> >> there's a bit of a very-long standing mess with the C xdr API.
>> >> >> >> >>
>> >> >> >> >> The problem is, the OTW XDR format for 32&  64 bit doubles is
>> >> >> >> >> identical on x86, amd64, sparc, and sparcv9.  In theory, the code
>> >> >> >> >> should then just merely copy the values out unchanged.
>> >> >> >> >> However, that really happened was on little-endian platforms, the byte
>> >> >> >> >> values for fp values were being (incorrectly) byteswapped by the
>> >> >> >> >> library.  Instead of fixing the library, it appears the existing
>> >> >> >> >> software that actually uses fp with xdr workaround the bug on those
>> >> >> >> >> platforms.  Of course the recent putback broke the workaround.
>> >> >> >> >>
>> >> >> >> >> Essentially:
>> >> >> >> >>
>> >> >> >> >> 1. code is released with broken xdr_{float,double}
>> >> >> >> >> 2. code is written w/ workarounds
>> >> >> >> >> 3. code is updated, and __amd64 define is (incorrectly) changed to
>> >> >> >> >> amd64.  This causes xdr_{float,double} to revert to a slower, platform
>> >> >> >> >> independent version that correctly (minus rouding errors) writes the
>> >> >> >> >> value into the stream (without swapping).
>> >> >> >> >> 4. My fix correctly writes fp values unchanged to XDR stream, but also
>> >> >> >> >> breaks any software with workarounds.  The test cases were written to
>> >> >> >> >> the RFC&  man pages (since I couldn't really find anything out there
>> >> >> >> >> that uses
>> >> >> >> >>
>> >> >> >> >> Given the pervasiveness of this, we probably need to do something similar
>> >> >> >> >> to:
>> >> >> >> >>
>> >> >> >> >> 1. Revert the behavior to the traditional broken behavior
>> >> >> >> >> 2. Update the man pages to document this behavior
>> >> >> >> >> 3. Add a #define to the rpc/xdr.h header file (XDR_FLOAT_NO_SWAP ?) to
>> >> >> >> >> allow applications to select the correct behavior when defined (and
>> >> >> >> >> document this in the manpage).
>> >> >> >> >>
>> >> >> >> >> _______________________________________________
>> >> >> >> >> Developer mailing list
>> >> >> >> >> Developer at lists.illumos.org
>> >> >> >> >> http://lists.illumos.org/m/listinfo/developer
>> >> >> >> >>
>> >> >> >> >
>> >> >> >> >
>> >> >> >> > _______________________________________________
>> >> >> >> > Developer mailing list
>> >> >> >> > Developer at lists.illumos.org
>> >> >> >> > http://lists.illumos.org/m/listinfo/developer
>> >> >> >> >
>> >> >> >
>> >> >> >
>> >> >> >
>> >> >
>> >> >
>> >> >
>> >
>> >
>> >
>
>
>

More information about the Developer mailing list