Re: 2 suggestions

[Richard Guenther <richard dot guenther at gmail dot com>]

On Apr 7, 2005 5:54 PM, David Edelsohn <dje@watson.ibm.com> wrote:
> >>>>> Ray Holme writes:
> 
> Ray> 2) Much of the time is spent in the several iterations of building a
> Ray> product doing the convfigure steps. These are repeated ad nauseum with the
> Ray> results being obtained the hard way each time. As a database person, it
> Ray> seems to me that by perhaps having a small database of configuration
> Ray> learned things (perhaps 3 key strings and one result string separated by a
> Ray> delimiter that awk recognizes) would allow the configuration to go much
> Ray> faster. I would be happy to write a small shell script that extracts
> Ray> information and another to add to such a database. The developer could
> Ray> then point  the make at where this default file is located using an
> Ray> environment variable - make and configure could use it and update it. Many
> Ray> others are using your style configurations and could benefit from such a
> Ray> database (gawk, gdb, .... as well as many others that copy your excellent
> Ray> style).
> 
>         As mentioned in another message, the user shell can contribute to
> a lot of the time.  Sometimes the default system shell is very inefficient
> and the configuration time could be improved by using GNU Bash.
> 
>         Also, the configuration process may look repetitive, but the
> results might be different in each situation, so GCC needs to inquire
> conservatively.

Also it helps a lot to remove paths to directories over the network (like NFS)
from PATH and LD_LIBRARY_PATH if this is possible.  Saves me half of
bootstrap time with our crappy network setup here at university.

Richard.