[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Running a plain ./config.status at top level

From: Alexandre Duret-Lutz
Subject: Re: Running a plain ./config.status at top level
Date: Thu, 10 Apr 2003 00:05:50 +0200
User-agent: Gnus/5.090016 (Oort Gnus v0.16) Emacs/21.2 (gnu/linux)

>>> "Akim" == Akim Demaille <address@hidden> writes:


 Akim> My latest proposal took this into account.

Sorry.  I answered old mail before popping new mail :(


 Akim> I thought about $? too, but I have no idea how portable that is :(
 Akim> That's why the latest patches did look at the most recent changed
 Akim> dependency.

That less accurate, though.  $? works at least on GNU, Solaris,
*BSD, OSF/Tru64, and HP-UX.  (It's POSIX, of course.)  Maybe we
could use $? just to discover whether it's really portable, and switch
back to `ls -1t' (or `ls -1dt', is the `d' important?) if it turns
out it doesn't work somewhere.  Otherwise we'll never know.

(BTW, `-1' isn't listed as a portable option of `ls' in the Autoconf
manual.  Is that an omission?)

 Akim> If you are concerned with speed, you might also consider adding
 Akim> an option to automake that would let it decide what
 Akim> are worth being reconsidered based on timestamps?  I don't know,
 Akim> wild quick and dirty guess before running to the restaurant :)

Coincidentally I just discovered yesterday, when working on the
fix for Jim, that there was already such code in Automake... :)
It's not enabled by default, you must invoke automake with
`--no-force'.  I see a few issues with the current

  - the timestamp check is done after the has been
    fully processed (just before is output -- or
    not), instead of right after the has been read.
    ( always needs to be read, to keep track of
    included files.)

  - it considers the time of wrt that of, aclocal.m4, and only: this is not
    enough once you start playing with `m4_include' in `', 
    or `include' in `'.  Fortunately the 1.8 branch 
    already knows all configure's dependencies (in @configure_deps), 
    and it should be easy to keep track of included files.

  - it doesn't touch Makefile.ins which are skipped.  From make's
    point of view this is quite bad.

Anyway, it's not clear where this option is useful.
It seems useless in Makefile rules, at least.  Maybe that
something `autoreconf' could use?  (I confess I always run
`autoreconf -f' so that's pointless in my case.)
Alexandre Duret-Lutz

reply via email to

[Prev in Thread] Current Thread [Next in Thread]