Bug#631572: Yes but no.

[Pierre Carrier <geoffroy.carrier@gmail.com>]

I really, really don't want upstream to be so stupid.
I might end up being completely wrong, in which case I apologize and
will buy a drink to author(s) of kconfig I offended.

I went to both extremities of my git repo, v2.6.12 and 3.0rc4, and I
found in both cases, in linux2.6.git/scripts/kconfig/confdata.c:
                if (p[0] == 'n') {
                        sym->def[def].tri = no;
                        sym->flags |= def_flags;
                        break;

But then again, I wanted you to be right, so I went back to
linux-2.0.tar.gz. In there, in linux/scripts/header.tk:

proc read_config { filename } {
	set file1 [open $filename r]
	clear_choices
	while { [gets $file1 line] >= 0} {
		if [regexp {([0-9A-Za-z_]+)=([ynm])} $line foo var value] {
			if { $value == "y" } then { set cmd "global $var; set $var 1" }
			if { $value == "n" } then { set cmd "global $var; set $var 0" }
			if { $value == "m" } then { set cmd "global $var; set $var 2" }
			eval $cmd
		}
		if [regexp {# ([0-9A-Za-z_]+) is not set} $line foo var] {
			set cmd "global $var; set $var 0"
			eval $cmd
		}


Back to Debian:

I'll admit that I barely understand this whole question, apart from
the counter-intuitiveness I experienced.
Please don't overreact if I'm talking out of my *ss.

- Would you therefore agree that declarations should have precedence
over comments?

- Do we have a good reason for this comment handling code in
kconfig.py is useful? (to me, those comments sound more like "we keep
the default for this one")

- If the answer to the previous 2 questions is "yes", but
implementation of precedence looks tricky/too automagical, can I offer
a patch that replaces all of those comments with "=n" to make the
overwriting of config made in lower-priority files more obvious?


-- 
Pierre