libtool-patches
[Top][All Lists]
Advanced

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

Re: generate package.m4 correctly on fresh checkout [libtool--gary--1.0-


From: Gary V. Vaughan
Subject: Re: generate package.m4 correctly on fresh checkout [libtool--gary--1.0--patch-1]
Date: Mon, 14 Mar 2005 12:09:53 +0000
User-agent: Mozilla Thunderbird 0.9 (X11/20041103)

Ralf Wildenhues wrote:
> Hi Gary,

Hallo Ralf,

Thanks for the reviews.

> * Gary V. Vaughan wrote on Fri, Mar 11, 2005 at 05:16:06PM CET:
> 
>>Okay to commit?
> 
> 
> I have a couple of questions and comments here: First, I'm uneasy with
> more and more stuff being built by `bootstrap' (which should ideally be
> empty or as small as possible).  Why can't we generate testsuite from
> `make'?  (It might just be me that forgot some point here.)

The golden rule is that we must not generate at configure time anything
that we want to distribute.  Until Alexandre pointed that out to me, the
bootstrap was indeed simpler, but getting `make distcheck' (or indeed any
VPATH build) to work involved an incredible amount of hair in make rules
to pick up things like `testsuite' from the correct place... and then when
an installer unpacked the resulting tarball with `testsuite' et.al. in
the source tree, more hair was required not to regenerate it in the build
tree.

> Failing that, why not create an empty stub and touching a prerequisite
> at the end so `make' will update package.m4?

We could do this, but I think an extra couple of lines in bootstrap is
cleaner than another timestamp file, and a bunch of make rule fiddling to
make sure some rules that we could just run directly from bootstrap are
rerun at the appropriate time (being careful to drop distributed files in
the source tree!).

> Aside, if bootstrap continues to grow (in size and importance), I think
> it should also be made more portable.  Like use of
>   : ${AUTORECONF=autoreconf}
>   : ${MAKE=make}
> 
>   env VAR1=val ... $MAKE tgts..
> 
> What do you think?

Yes, that seems like a good plan to me.  I imagine that we have gotten away
with it so far merely because people who track CVS have a full set of gnu
utilities first in their path.

> And finally, returning to the patch: I could not reproduce the bootstrap
> failure on linux/GNU make.  Where does it fail?

  $ cvs -d `cat libtool-HEAD/CVS/Root` co libtool
  ...
  $ cd libtool
  $ ./bootstrap
  ...
  cd .; \
  rm -f tests/defs.in; \
  autom4te --language=m4sh -B ./config tests/defs.m4sh > tests/defs.in
  make: *** No rule to make target `testsuite'.  Stop.
  ...

Cheers,
        Gary.
-- 
Gary V. Vaughan      ())_.  address@hidden,gnu.org}
Research Scientist   ( '/   http://tkd.kicks-ass.net
GNU Hacker           / )=   http://www.gnu.org/software/libtool
Technical Author   `(_~)_   http://sources.redhat.com/autobook

Attachment: signature.asc
Description: OpenPGP digital signature


reply via email to

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