Re: [gentoo-dev] portage-2.0.51_pre10: QA Notice Messages

[Brian Harring]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


On Jun 11, 2004, at 2:01 PM, Ned Ludd wrote:

> On Fri, 2004-06-11 at 14:17, Paul de Vrieze wrote:
>>> IUSE isnt defined when portage hits that (it is defined later on) so
>>> you
>>> get the warning message ... portage needs a way to handle conditional
>>> inherits, so you cant really say the eclass/ebuild is broken
>>>
>> No it doesn't, those eclasses / ebuilds need to be written in such
>> that there
>> is no need for a conditional use of them.
>
> Well in the case if the flag-o-matic one, it's going to yap at us no
> matter what. The IUSE is defined unconditionally, first properly
> IUSE=debug then later on IUSE="$IUSE debug"
>
> The only way I've found to get portage to shutup and behave half sane
> is
> to hack the .ebuild.sh
>
> -if ! hasq "${u}" ${IUSE} && ! hasq "${u}" ${PORTAGE_ARCHLIST} selinux;
> then
> +if ! hasq "${u}" ${IUSE} && ! hasq "${u}" ${PORTAGE_ARCHLIST} selinux
> debug; then
Use this instead;

if ! hasq "${u}" ${IUSE} && ! hasq "${u}" ${PORTAGE_ARCHLIST} selinux; 
then
- -	echo "QA Notice: USE Flag '${u}' not in IUSE for ${CATEGORY}/${PF}" 
 >&2
+	if ! hasq "${u}" ${E_IUSE}; then
+		echo "QA Notice: USE Flag '${u}' not in IUSE for ${CATEGORY}/${PF}" 
 >&2
+	else
+		echo "QA Notice: USE Flag '${u}' was accessed in the global scope, 
but is not available until after the ebuild/eclass has been fully 
sourced: ${CATEGORY}/${PF}" >&2
+		echo "            the offending function/check probably belongs in 
pkg_setup()." >&2
+	fi
fi


For .51 all iuse definitions by eclass's are shoved into E_IUSE, to 
protect them from being stomped on by A) other eclasses (an eclass, 
inheriting another eclass for instance), and B) the ebuild itself.
Using glibc as an example, ebuild.sh does the inherit's, pushing the 
eclass defined IUSE's into E_IUSE; after the ebuild has been sourced, 
E_IUSE is merged into IUSE.

Problem is that some ebuilds have calls in the global scope that access 
those vars- strip-flags in glibc for instance.
Alternatively, could modify useq so that it checks both IUSE and 
E_IUSE, rather then what I defined above...
~brian
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (Darwin)

iD8DBQFAyi9EvdBxRoA3VU0RAkZrAKDcnA1ztPiXI53k64dXcwJANkgG2gCfdffk
3dhZFAHoHe4XnOdtadYVkRw=
=djQn
-----END PGP SIGNATURE-----


--
gentoo-dev@g.o mailing list