This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Bootstrap Comparison Weirdness
- From: Jim Wilson <wilson at specifixinc dot com>
- To: Ranjit Mathew <rmathew at gmail dot com>
- Cc: gcc at gcc dot gnu dot org
- Date: Tue, 06 Jul 2004 13:19:14 -0700
- Subject: Re: Bootstrap Comparison Weirdness
- References: <cc18qd$c3$1@sea.gmane.org>
Ranjit Mathew wrote:
I've noticed this quite a few times but only sometimes
can I reproduce this cleanly. What happens is:
1. Bootstrap from clean sources in a clean folder.
2. Change something in a front-end (say Java).
3. Rebootstrap. Everything goes through fine.
4. Rebootstrap without changing anything, but now I get
bootstrap comparison failures for files touched in #2.
This is trivial to reproduce, and it isn't hard to see why. What I did
was a C only bootstrap, and then I added "int phony = 100;" to the end
of dwarf2out.c.
What happened is that a top level bootstrap calls the gcc bootstrap
rule, which has no dependencies on the source code. This emits a
helpful message saying that bootstrap is finished, and you probably
meant to do something else. The top level bootstrap rule then does a
file compare which passes because nothing changed. The top level
bootstrap then tries to build the libraries. But some of the libraries
have a dependence on gcc, which depends on its source files, so now cc1
gets rebuilt. Meanwhile, the helpful message from the gcc bootstrap
rule gets lost in the make noise.
Next time you do a bootstrap, you get the same sequence, except the file
compare fails this time because files have changed.
This seems to be a design conflict between the top level and gcc
bootstrap rules. What the gcc bootstrap rule is doing makes sense if
bootstrap is completely controlled by gcc, but it does not seem to make
sense when the bootstrap process is controlled by the toplevel Makefile.
I'd suggest filing a bugzilla bug report if you haven't already.
--
Jim Wilson, GNU Tools Support, http://www.SpecifixInc.com