Re: lynx-dev removing strings used in CTRACE when configured without tracing

[Vlad Harchev]

On Sun, 18 Apr 1999 address@hidden wrote:

> > The use of '_' instead of ',' needs getting used to, but apparent it gets 
> > around the variable-argument-number problem nicely. 
> 
> maybe - though I'd have to test & see that the different preprocessor
> agree (some people tend to think that if it works in gcc, it's portable -
> I can remember an order-of-evaluation problem in gcc which the GNU people
> had to abandon 4-5 years ago because it wasn't ANSI compatible)
>   
> > It would be a problem if we used _() as a shhorthand for gettext(); but 
> > fortunately the lynx code uses gettext() directly. 
> >  
> > This would require replacing all TRACE messages throughout the code; 
> > but so would the other approaches that are worth considering, I think. 
> >  
> > I agree with Tom though that this is probably not the best time for it. 
> 
> yes - I'd rather do stuff like that early in a cycle.  This week I'm working
> to finish off changes for xterm for a code freeze in XFree86, then (unless
> people have different ideas) it's time to do the same for lynx.
>[...]

  One of the solutions may be the following:
 There is a standalone preprocessor distributed with gcc, it's called 'cccp'
and it does support macros with variable number of args. 
 It's self contained C file 300 KB long + cccp.1 15KB long. We can include it
in the distribution of lynx, so it will be compiled by make before other
sources of lynx (as gettext if absent), and be used for all lynx sources.
'cccp' is in the 'gcc' subdir of gcc distribution. 
 In this case we can use macros with variable number of arguments. This won't
require touching all *.c files.

 BTW: cccp binary is not shipped with RedHat 5.2, tho' cccp.1 is.

 Best regards,
  -Vlad