[PATCH] Bulletproofing -fsyntax-only, round 2 (extra armour plate)

Nix nix@esperi.demon.co.uk
Wed Jun 14 12:40:00 GMT 2000

Nix <nix@esperi.demon.co.uk> writes:

> Here's a patch with both these changes:

... and, naturally, a stupid error, so small it's not even worth rolling
a patch for.

> diff -uprN egcs/gcc/system.h egcs-hacking/gcc/system.h
> --- egcs/gcc/system.h	Sun May 28 23:18:17 2000
> +++ egcs-hacking/gcc/system.h	Sun Jun 11 21:59:12 2000
> @@ -589,6 +589,13 @@ extern void abort PARAMS ((void));
>  #define ONLY_INT_FIELDS 0
>  #endif 
> +/* Provide a default for the HOST_BIT_BUCKET.
> +   This suffices for POSIX-like hosts.  */
> +
> +#ifndef HOST_BIT_BUCKET
> +#define HOST_BIT_BUCKET "/dev/nill"
> +#endif
> +  
>  /* Enumerated bitfields are safe to use unless we've been explictly told
>     otherwise or if they are signed. */

*That*, of course, should be "/dev/null", not "/dev/nill".

I typoed and the patch was made against one of my test trees --- the one
where I intentionally misspelled the name of the bit bucket to force the
fallback-to-temporary-file code to be used. This patch'll still *work*,
but always using a temporary file when there's code there to use a bit
bucket instead is rather braindead :)

It's true --- on the net, you *can* make a fool of yourself in front of
everyone... :(

(btw, Zack, in case you care, I've backported your implicit-declaration-
of-builtins warning patch to 2.95.2. Available on request.)

> ... knowing the alignment of Orcs in AD&D.
  --- David Jacoby and Greg Andrews in the Monastery

More information about the Gcc-patches mailing list