Re: [PATCH]: random global function name

[Paul Schlie <schlie at comcast dot net>]

WRT #import implementation:

Personally can't imaging how it could be considered acceptable to rely on a
hash of any sort to be used as a basis of determining equivalence, when by
its very definition will be known to unpredictably yield false equivalence
predictions.

Further more, it's not clear to me that it is even semantically desirable
to exclude explicitly coded multiple #import of unique files which may in
fact have literally equivalent content, which a hash based equivalence
mechanism would prevent; as it does not seem likely consistent with the
intent of the language feature.

Lastly, I'd hate to be in the position of trying diagnosing an insidious
#import failure of either of the two aforementioned types, in either a
stable or non-stable source code base, caused by a hash based equivalence
prediction failure.

So in summary, although I appreciate that defining file equivalence based on
the lexical comparison of a files absolute resolved path may not be perfect,
it certainly seems to be more clearly and less ambiguously definable,
implement-able, controllable, and debug-able; for what that may be worth.

-paul-

> On Thu, 10 Jul 2003 17:13:09 -0700 , Mike Stump wrote:
>
>> On Thursday, July 10, 2003, at 04:41 PM, Richard Henderson wrote:
>>
>> I will grant that the use of a equivalent here in gcc is less evil than
>> elsewhere, since we're hashing something that is already known to
>> be non-unique (and so it doesn't really change the outcome much),
>> but I still object because there was no explanation at all as to
>> why the change was supposed to be worthwhile.
>
> Geoff plans to make #import work by hash.  Speak now.  :-)