Sender: current-users-owner@NetBSD.org

Greetings everyone,


I have a pair of questions.  Has anyone taken a look at all the
warnings which are being raised during a build of 2.0, at least on the
i386 platform?  And is there any chance some, if not a significant
portion of them may be resolved before we complete the 2.0 release
cycle?  

In case you do not know what I am talking about, I just completed a
second build within the past week of 2.0/i386 and found over 51000
warning messages in the build log.  This latest build is from sources
obtained from CVS as of Aug 13 00:00 ET, but is not unique to this
build.  Instead, I saw it even in my first build on the 2.0 branch
back in mid-April, and at the time I just wrote it off without
research as the result of some ongoing change to either GCC or the
compiler flags, and have pretty much forgotten about it until today.
Indeed, I had thought to re-enable some not-yet-commited experimental
code which reported the number of warnings/errors, but decided to look
at the results first to see what was in store for me.

Prior to writing this message, I did a quick check of my archived
messages for all of the NetBSD mailing lists to which I am subscribed.
I am somewhat surprised that I did not find mention of these warnings,
especially given how they amount to over 15% of the total output.  Is
anyone else seeing these messages?  Now, for a comparison, a full
build of 1.6.2 only gave me 7838 warnings last time I did it back on
Aug 4, which was about 6% of the total log.

What are the error messages?  Well, I still have one build going, but
it does not look appreciably different from the build I did
Monday/Tuesday.  In that build, over 17K of them are warnings
about "pointer casts may be troublesome".  Another 6200 are "possible
pointer alignment problem", ~5500 are "suggest not using #elif in
traditional C" and so on.  A full summary of the error messages can be
found at the following URL:

    http://cinnion.ka8zrt.com/netbsd/warns.0

The associated build log for these warnings is located at:

    http://cinnion.ka8zrt.com/netbsd/do_build.log.0.gz

It shows the date of the build, as well as the dates for the
associated CVS updates.

The warnings were extracted with the command line:

    gzcat do_build.log.0.gz | sed -ne 's/s/.* warning: /warning: p' | sort | uniq -c | sort -rn

FWIW, these builds are done in a sandbox, with just a handfull of mods
to the source so that I do not have to update my host file, and so
that a few additional items such as customized kernels are
automatically built.  You can see the build scripts (the one of
particular interest here is do_build.sh) at:

    http://www.ka8zrt.com/cgi-bin/cvsweb.cgi/netbsd_build/?only_with_tag=netbsd-2-0-0

And the output from my CVS update is at:

    http://cinnion.ka8zrt.com/netbsd/update.log.0

Looking forward to seeing what other folks have to say about this
situation.

- Doug

-- 
Douglas Wade Needham - KA8ZRT        UN*X Consultant & UW/BSD kernel programmer
Email:  cinnion @ ka8zrt . com       http://cinnion.ka8zrt.com
Disclaimer: My opinions are my own.  Since I don't want them, why
            should my employer, or anybody else for that matter!