-Wall and all that

[Kai Henningsen]

jsm28@cam.ac.uk (Joseph S. Myers)  wrote on 16.03.02 in <Pine.LNX.4.33.0203162333530.19686-100000@kern.srcf.societies.cam.ac.uk>:

> This idea was extensively discussed in September 1998; there was even a
> patch by Mark Mitchell that was checked in for less than a day.  Reading
> the previous discussion from egcs and egcs-patches is recommended.  There
> was no consensus support for message identifiers.  gettext works fine for
> i18n.

(I already addressed the i18n part.)

I have now read that thread. I must say I am appalled.

The most extensive part to the argument boils down to "let's keep the  
users' life hard, it's good for them". I won't even try to comment on that  
part; I don't think I could do it without using abusive language.

If I subtract that part, what is left is a clear general consensus that  
warning control is good, that using pure numbers is bad, and that it  
should wait for the gcc/egcs merge. I agree with all of them (well, the  
last goes in by merit of having now happened), and none conflict with my  
description.

As for the "how hard is it to filter the output with sed", well, I  
actually did that once; I dropped that long before making it part of a  
makefile, because I considered it an entirely unreasonable way of going  
about things.

As for the "silence system header warnings" part, most of my experience at  
the moment is still with 2.95.3 where, frankly, this is utterly broken.  
However, I don't consider this as related to the warning control  
discussion - this part really should be solved by just handling system  
headers better. Maybe 3.x already does that. I've had too many  
instabilities (not related to gcc) in the relevant product lately to even  
consider a compiler upgrade until this settles down; one change at a time.

MfG Kai