There's a line that's already in pygobject' configure.ac, that's also in pygtk git, but not in portage, that's needed for a fully correct configure run, namely m4_ifdef([LT_OUTPUT], [LT_OUTPUT]) I think I'll file a bug in gnome bugzilla shortly, cause I disagree with this "solution" (made in pygobject and pygtk). Well, there is a chance, they're doing the right thing, but not, if I understood certain thread on libtool list many months ago. That thread said that even as early as libtool 1.4 there was a possibility to check for those vars directly.
but what problem is it causing ?
Frankly, I'm not sure if it makes a difference. This is the check they're doing: dnl get rid of the -export-dynamic stuff from the configure flags ... export_dynamic=`(./libtool --config; echo eval echo \\$export_dynamic_flag_spec) | sh` Later on, they sed out what got assigned to this var (IIRC, on Linux, with gcc it's -Wl,--export-dynamic). Maybe they do have a reason for this. And as I mention, they've "fixed" in that way. Probably "fixing" this here is not that important, given that any new release will have their fix, but I thought it's worth mentioning anyway.
doesn't sound too big of an issue to require changes in the ebuild. Please check this with upstream directly.