Re: [Mingw-users] Fixing UNICODE related macros in w32api headers

[Christopher F. <cg...@re...>]

On Tue, Sep 10, 2002 at 11:53:46PM -0400, Igor Mikolic-Torreira wrote:
>At 11:29 PM 9/10/2002 -0400, Christopher Faylor wrote:
>>>I can solve is by adding "__attribute__((always_inline))" to the
>>>functions.  This way so there are no "extra overhead" concerns even in the
>>>no optimization case.
>>
>>Is always_inline available to gcc 2.95.x?  I thought this was a very
>>recent addition to gcc.
>>
>>Checking the gcc ChangeLog...
>>
>>Yes, it was introduced on 2002-02-21.  So, it's in 3.2 but not 2.95.3.
>>
>>Sorry.
>
>Bummer.  Well, is the overhead of having these functions out-of-line in the 
>no-optimization case a show-stopper?  The answer is probably "depends" ;)

I would really like to see everything defined as inline rather than as
a macro myself but I think "it depends' is the right answer.  Not that
I have any say in this.  :-)

You could define things differently based on whether you are compiling
for gcc 3.2 or greater but that would probably needlessly complicate the
header files.  Keeping the headers compact is a goal of w32api, AFAIK.

>I'm still going to make these mods locally for our own work-group because 
>it makes our lives many orders of magnitude easier (we do a lot of porting 
>& adapting) and the extra overhead in the no-optimization case isn't a big 
>concern to us.  So for _our_ situation it's pretty clear-cut.
>
>I'm not sure what everyone else will think.  If the consensus opinion is 
>that this change is worth the cost incurred in the no-optimization case, 
>let me know and I'll submit a patch.  I am most definitely not trying to 
>force this change on everyone -- just offering to share if so desired.

It should be interesting to hear what everyone else thinks about this...

cgf