Re: Pretest

[Eli Zaretskii]

> From: Chong Yidong <address@hidden>
> Date: Sat, 28 Oct 2006 16:47:07 -0400
> Cc: address@hidden
> 
> I'm using the version shipped with Ubuntu Dapper, version 2.59a-7

This version sounds like it's some kind of development snapshot.  Can
you try to find out?  Maybe we shouldn't be using this version.

> We could revert configure to the last version used, and stick with
> the configure script for the rest of the pretest.

That's not a good solution, IMO, since some bugs reported during the
pretest might require to regenerate the script.

> > I'd still like to understand what failed `ld'.
> 
> I think the easiest thing to do is to apply the hunks of the diff
> until you find which part causes compilation to fail for you.  It
> shouldn't take that long with a binary-search.

I don't know when I'll have time to do that; fencepost.gnu.org is not
my main machine.  Even if I do find the portion of the diffs that
causes the problem, then what?  We don't have a version of Autoconf
that would have the effect of applying only part of the diffs, so once
again we will be in a position where regenerating the configure script
would be difficult of impossible.

And on top of that, I still have trouble understanding how any change
in the configure script _alone_ could have anything to do with this.
Unless I'm missing something, there are only 2 ways Autoconf can
affect the temacs build: either through config.in/config.h, or through
src/Makefile.  These are the only two things I know off that affect
compilation and linking of the C sources.

However, in this case, config.in and config.h are identical to the
ones you included in the tarball, and so is src/Makefile.in.  Am I
missing something?

So I'm still thinking the `ld' failure was some real problem, and the
changes in the configure script are only indirectly related to it.  If
someone knows how to investigate the `ld' failure (short of rebuilding
Binutils with debug info and running `ld' under a debugger), I'd
appreciate any ideas.