LAM/MPI logo

LAM/MPI General User's Mailing List Archives

  |   Home   |   Download   |   Documentation   |   FAQ   |   all just in this list

From: Ralf Wildenhues (Ralf.Wildenhues_at_[hidden])
Date: 2004-09-24 02:35:25


Jeff Squyres writes:
> On Sep 22, 2004, at 3:59 AM, Ralf Wildenhues wrote:
>
>>> I have seen builds get out of whack like this a few times; I think it's
>>> been due to situations like: - when you run configure multiple times and
>>> do not run "make clean" before re-running "make"
>>
>> To avoid this needs very careful source layout (Makefile.am mainly).
>
> AFAIK, we have such a layout. I rarely run into these errors (have not
> for some time now) -- this is mainly based on anecdotal experiences from a
> long time ago (probably with prior versions of AM) and where I did not do
> a complete analysis of what the exact problem was. In most (all?) cases,
> a "make clean all" fixed things, however.

Which defeats the use of a Makefile, mostly.

>>> - Solaris' dependency analysis gets mucked up upon successive
>>> invocations of "make"
>>
>> This sounds like a bug in `depcomp'. Care to write a bugreport against
>> Automake?
>
> Not to worry -- I'm not shy there (I have submitted a bug or two against
> AM). This is anecdotal recollections, however, so I have nowhere near the
> appropriate data for a proper bug report.

OK.

> Another important thing to check (sorry -- should have thought of this the
> first time I posted) is that if you're building on an NFS-mounted
> filesystem, ensure that the time on your client machine is tightly
> synchronized with that of your NFS server (e.g., with ntp).

There might be more trouble than that. Recent linux systems have higher
resolution time stamps in certain cases, on certain filesystems. I think
this has led to all sorts of interesting problems. If in doubt, keep
source and build on the same, local filesystem.

Regards,
Ralf