[EMAIL PROTECTED] writes:
> Hey guys,
>
> I have made a number of modifications to NLBConfig, listed below. If
> you consider these changes to be useful to your project, let me know
> and I may do some more, based on your feedback. The overall reasons I
> made these changes were:
>
> - To make the Makefile more readable and better documented, e.g. where
> do the different parts of the Makefile come from, what do they do,
> etc.? This helped me understand the build process a lot better, and I
> think it would help other newcomers also. Maybe oldtimers as well,
> when you experiment with new configurations, as you have done a lot
> lately.
This sounds like a good goal. Even if we automatically generate it,
having a readable config file is appreciated.
> - To make NLBConfig a bit more readable, behave more consistently,
> give more informative diagnostics, be a little easier to use.
>
> I have tried to maintain compatibility with existing config files, and
> as far as I know I haven't broken anything, but I have only tested it
> with a SiS630 build (and I don't have the hardware to test those bits
> with).
>
> ...also, a few other notes/suggestions...
Ah so we need to be carefull :)
> - In the data sheets for a couple EPROMS and FLASH parts that I looked at,
> erasing the part results in memory being set to 0xFF. And I've noticed
> with my EPROM programmer that when unused bytes are filled with FF (instead
> of 00), the programming is much faster. So, you might consider changing
> mkrom and similar tools, to have them set unused bytes to FF. It might
> speed up the FLASH programming.
Good suggestion.
> - This code in ~/freebios/src/arch/i386/config/make.base looks a bit odd:
>
> addaction linuxbios.rom dd if=linuxbios.strip of=linuxbios.rom bs=1 seek=`expr
> $(ROM_IMAGE_SIZE) - $$size`
>
>
> For one thing, "dd ... seek" doesn't pad out the beginning of the file
> if the file doesn't already exist, at least on my system.
It does on mine, and that was the intended behavior. Are you certain
you had a ROM_IMAGE_SIZE greater than your image start?
> Instead, it
> just acts like cp - which I think is ok, because mkrom takes care of
> padding and so on.
mkrom is no longer run. At least it shouldn't be.
> If I've misunderstood something and you really want
> to pad the beginning of the file (with 0xFF) you can do the following:
>
> addaction linuxbios.rom perl -e 'print "\377" x ('$(ROM_IMAGE_SIZE)-$$size')' >
> $@
> addaction linuxbios.rom cat linuxbios.strip >> $@
Since I've introduced perl in a few other places that looks like a much
better action.
> - There are several files in the freebios/src tree that have lines of the form
> static char rcsid[] =
> "$Id: northbridge.c,v 1.8 2001/11/06 04:29:38 ollie Exp $";
>
> These files cause "defined but not used" warnings when compiled. The
> warnings are just noise that may be drowning out possibly more serious
> warnings. The "not used" warnings can be cured by just changing the
> first line to
>
> static char rcsid[] __attribute__ ((unused)) =
That sounds good. It would probably do even better to make a
macro RCSID. i.e.
#define RCSID(x) static char rcsid[] __attribute__ ((unused)) = x
And then we could be easily consistent.
>
>
> Cheers!
> - Jan
>
>
> The changes to NLBConfig are as follows:
>
> - Put more whitespace and comments into the Makefile.
>
> - The "include cpuflags" in the Makefile was broken: namely, the
> cpuflags file was generated long after it was needed. So I eliminated
> the cpuflags file and used a different mechanism to compute CPUFLAGS.
> Yet another way to do this would be to just have NLBConfig write
> "CPUFLAGS := ..." in Makefile.settings.
Was it really broken. It works here and I haven't heard complaints.
GNU make is pretty smart about building makefile depenencies.
Ideally I'd like to have 2 tools.
Build-Makefile
Build-Configuration
And with everything being derived from Makefile.settings and regenerated
we don't have a problem seperating things out.
> - NLBConfig.py was renamed to NLBConfig and made executable. So you
> just say "NLBConfig ... ..." rather than "python NLBConfig.py ... ..."
>
> - The build directory is created if necessary. This eliminates a
> dozen lines from the HOWTO document.
Quite reasonable.
> - The "Trying to create" messages were replaced with "Creating".
> Let's have a confident attitude :-)
>
> - Deleted some apparently obsolete comments, rewrote some code to
> hopefully improve clarity. Removed unreachable feature "list_vals".
>
> - During make, the gcc lines are printed as "gcc ... -o nsuperio.o
> nsuperio.c" in one line rather than several lines of repetitious and
> noisy details.
I'll have to look but that sounds good. Having all of the defines
on the command line is probably a bad idea but it was a good first stab.
> - Renamed some variables in the NLBConfig script to better match the
> command with which they are associated, for example outputdir is
> renamed target_dir.
>
> - There seemed to be no reason to have separate userrules and
> makebaserules, so userrules was eliminated.
>
> - Changed the rules for remaking Makefile when config dependencies
> change so as to avoid running NLBConfig twice.
Hmm. I haven't seen that but it sounds o.k.
> - Improved error detection and reporting, in particular, report the
> config file in which any duplicate makerule target definition occurs,
> and in which command any syntax error occurs; detect and report more
> syntax errors.
>
> - Cause makerules to be written to Makefile in the order that they are
> defined, rather than hash order.
>
> - Change the parser for makerule and addaction to allow "#" comments
> to be passed to the Makefile as actions. This allows the comments to
> be printed when make runs.
>
> - Makerule remembers the config file where the makerule was given, and
> writes it in the Makefile, so it's easier to tell where rules come
> from.
Cool.
And you missed the number one detail on my todo list. Add header file
dependencies so we have complete dependency information.
Eric
