[ On Friday, February 25, 2000 at 16:49:42 (-0500), Harlan Stenn wrote: ]
> Subject: Re: config.cache considered harmful
>
> I'd almost like to see a "version" number in the cache, which I can "bump"
> whenever I make a "significant" change to the logic which has the effect
> of...
It has always been a very good idea to put version identification
information into persistent (or shared) data. If everyone did this
uniformly in every software product the world would probably have at
least an order of magnitude less bugs.... The only problem with doing
this is in knowing when it's important to change the version
identification (which obviously in this case includes times when the
information is being used on a "different" platform, whatever that might
mean! :-).
I too am strongly in favour of disabling "config.cache" support by
default. While it may be extremely helpful in a small number of cases
as a performance enhancement it is dangerous in a wide variety of
situations covering both developers and end users.
In fact I would go so far as to suggest that "config.cache" is only
truly safe in its current form when it is used for nested configure
scripts, and then only when the nested scripts are all directly realted
in heritage to each other. Until version information (including much
more platform details than just what config.guess reveals) is included
in the cache file, and carefully checked by every consumer,
"config.cache" shouldn't even be allowed to be turned on.
Even when version information is included in "config.cache" I think that
"autoconf" should remove it unconditionally, as should any makefile rule
which is about to run "configure" but detects that the script is newer
than the cache file.
--
Greg A. Woods
+1 416 218-0098 VE3TCP <[EMAIL PROTECTED]> <robohack!woods>
Planix, Inc. <[EMAIL PROTECTED]>; Secrets of the Weird <[EMAIL PROTECTED]>
