Build error 2.3 on Ubuntu lucid beta

[Colin Law]

On 25 March 2010 20:35, Derek Atkins <warlord at mit.edu> wrote:
> Hi,
>
> Please remember to CC to list on all replies..

Sorry, this list is always catching me out, other lists I subscribe to
just use Reply for reply to list, and members tend to complain if
reply to all is used.

>
> Colin Law <clanlaw at googlemail.com> writes:
>
>> On 25 March 2010 18:08, Derek Atkins <warlord at mit.edu> wrote:
>>> Colin Law <clanlaw at googlemail.com> writes:
>>>
>>>> On 25 March 2010 15:31, John Ralls <jralls at ceridwen.us> wrote:
>>>>>..
>>>>> You need to run autogen.sh if a Makefile.am is changed. Autogen runs automake which makes Makefile.in from Makefile.am; configure uses Makefile.in, not Makefile.am, for input to create Makefile.
>>>>
>>>> Would it be possible to have a top level make that ran autogen and
>>>> configure when appropriate?  One would have to provide the params for
>>>> configure somehow of course.
>>>
>>> Generally the auto-tools will re-run it for you in maintainer mode.
>>> I honestly don't know why it didn't re-run for you.  It certainly does
>>> it for me on Linux.
>>
>> Sorry I don't know what you mean by that, could you explain in more
>> detail?  I can't see how configure can run automatically at the moment
>> as you have to supply parameters.
>
> No, you don't.  config.status knows how you called configure.  C.f.:
>
>  ./config.status --version
>
> The auto tools are set up such that it should notice that the
> Makefile.am is newer than the Makefile.in and rerun automake, or that
> the configure.in is newer than configure and rerun autoconf.  It's
> supposed to all just work (and it works just fine for me here on Linux).
>
> In the case of rerunning configure, it just calls ./config.status --recheck

Curiouser and curiouser.  I have just checked that if I edit the
offending Makefile.am and run make it does regenerate Makefile.in and
Makefile correctly, and if I revert the change and get the compile to
fail, then restore the change and it compiles ok.  I can only assume
there was something strange about the timestamps on the make files
that prevented them from being regenerated.  I notice that make clean
does not remove them which presumably it could.  I have checked that
other files that I picked up when I updated my git-svn repository have
the correct timestamps.  Very odd, and now of course the evidence is
gone, so it may remain one of the great unsolved mysteries.

Colin