This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: egcs fixinc: This Time For Sure (tm)
- To: mrs at wrs dot com (Mike Stump)
- Subject: Re: egcs fixinc: This Time For Sure (tm)
- From: Jeffrey A Law <law at cygnus dot com>
- Date: Thu, 05 Mar 1998 00:45:30 -0700
- cc: egcs at cygnus dot com, korbb at datadesign dot com
- Reply-To: law at cygnus dot com
In message <199803050127.RAA24111@kankakee.wrs.com>you write:
> Sometimes when I put new dangerous code in the compiler, I leave the
> old stuff in, and run the new stuff in parallel, and compare the
> results. I then AI match the results, and use the old stuff if they
> differ in ways I don't expect. If they differ in ways I expect or are
> the same, I use the new stuff.
>
> After a suitable time span, I then rip out all the old code.
>
> (Yes folks, we can practice software fault tolerance. :-))
>
> Fixincludes sounds like one of those things were doing this would be
> fairly easy. Also, once checked you can add the triplet to a list of
> known good systems (to regain the speed) to mitigate the slowdown
> doing this would otherwise cause.
Well, it's an interesting issue.
Part of me wants to take a very cautious approach to this conversion.
Part of me also wants to avoid dual maintenance if at all possible
and to get through the the conversion process as quickly as possible.
It's not clear which will win :-)
It may be the case that we have Bruce's new stuff available for
1.1, but off by default. Then switch it to be the default and
remove the old crap for 1.2. Dunno.....
jeff