On Thu, Jul 26, 2012 at 2:49 PM, Jed Brown <[email protected]> wrote:

On Thu, Jul 26, 2012 at 2:44 PM, Matthew Knepley <[email protected]>wrote:

...

You do not read it correctly. It operates exactly as everything else. You move make.log to make.log.bkp, which is what we do now, and write into make.log. It seems to be writing without moving first.

...

I think it is wrong to reuse the name make.log for both purposes. Regardless, the symlinks are not removed, so make.log output overwrites $PETSC_ARCH/conf/make.log and (the next time you run it) $PETSC_ARCH/conf/make.log.bkp.

In my first message, it says "this is a bug". What is hard about that? What necessitates 3 replies missing that point?

Sounds like your "correct behavior" is exactly what I'm complaining about. Since there is only one level of backup, running twice, as in

./builder2.py buildExample src/... ./builder2.py buildExample src/...

destroys all trace of the make.log for the library.

How can your make.log go away if its in PETSC_ARCH/conf ? Is this a useful argument? Matt -- What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead. -- Norbert Wiener