maintainer mode

[Alexandre Oliva]

On Oct 31, 1998, Andrew Tridgell <tridge at samba.anu.edu.au> wrote:

> it will break, but only on systems that can't handle -c and -o on the
> same command line. What kind of system will have a parallel make and
> not be able to do that? 

People usually build GNU make before gcc, because gcc requires GNU
make to build :-)

Anyway, now I clearly see your point.  You're thinking as a developer
of a particular package (or set of packages), so you don't care so
much about *general* solutions.  I, as a member of the
autoconf/automake/libtool teams, *do* care and *must* care about
*general*, *foolproof* solutions, otherwise novices will blame on the
developer of the package, and he will blame on us :-)

So we must be very careful about what we can give up on supporting.
You don't have to.  And this is fine.

> then just forget the locking. If someone tries a parallel make on an
> ancient SCO system that can't do -c and -o then they are very
> silly.

The point is: how long will it take for them to figure out that they
must remove foo/x.o and rebuild without -j, because that file should
have been named bar/x.o?  How many bug reports will be forwarded from
random packages to automake at gnu.org or bug-libtool at gnu.org before we
figure out what the problem was, i.e., that he was running make 0.01
and a broken port of gcc 0.55 he had done himself to his Casio Digital
Diary? :-D

> the fact that it actually will work because samba doesn't have any C
> files of the same name in 2 directories is largely irrelevant :-)

:-)


BTW, there are two annoying things in the mailing lists of Samba:

1) it removes all Cc:, so automake people are no longer listening to
us

2) it removes other headers such as the essential ones to decode MIME
messages.  Not funny to get tar.gz base64-encoded bug reports this way 
:-)

-- 
Alexandre Oliva
mailto:oliva at dcc.unicamp.br mailto:oliva at gnu.org mailto:aoliva at acm.org
http://www.dcc.unicamp.br/~oliva
Universidade Estadual de Campinas, SP, Brasil