Re: configure warning suggests an anti-pattern

[Bruno Haible]

Thomas Dickey wrote:
> It's obviously too late for config.guess to be improved by taking into
> account the standard way for specifying compiler options (in $CFLAGS),
> since that would immediately break programs which are being configured
> by the -Werror clique.

Yes, the behaviour of config.guess is essentially not modifyable
without breaking many GNU packages.

Ben Elliston, when I talked to him about the use of $CC for bi-arch
determination a couple of years ago, essentially said "I wish it wasn't
there; I would like to get rid of it but I can't; let's not make things
worse by considering also $CFLAGS and what else".

> overbroad, certainly.  But taking this into account:
> 
>       https://gcc.gnu.org/onlinedocs/gcc/Preprocessor-Options.html
> 
> and this:
> 
>       https://gcc.gnu.org/onlinedocs/gcc/Invoking-GCC.html
> 
> it's not something that I want to spend a lot of time, tuning
> things to work around nonstandard options which have numerous
> inconsistencies which preclude a naming convention.

I agree that reshuffling options is hard, because each compiler has
their own big set of options.

> If you happen to notice a case where the program fails to configure

ncurses builds fine on the platforms on which I've built it so far.
Kudos!

FYI, here are the values of CC that I recommend on various platforms:
https://gitlab.com/ghwiki/gnow-how/-/wikis/Platforms/Configuration
Feel free to contribute if you have access to platforms that are not
listed yet.

Bruno