i should probably finish porting the interface_check script to C (in the spirit of speeding up builds)...

just how pvs handles some of the verdef info to produce the output is a bit hard to follow and reproduce

and (AFAICT) not actually documented

okay, I think I have something working for me.

sommerfeld: sorry it was so rough

okay, really figured out the problem -- residual config.status file in the git tree that caused one of the sub-configures to emit `configure:3813: error: source directory already configured; run "make distclean" there first`

uugh

trying to debug automake configure scripts is its own circle of hell

jbk: the GCC ones are particularly special (I think, perhaps, they're the genesis of auto*)

gcc configures (top level, some sub levels(?)), and then configures other bits under 'make' too, so if the later ones fail they're desparately interleaved in stderr

we have a product that combines some management bits with openzfs + lustre + infiniband on Linux.... I think it's one of the lustre bits that has a 40,000+ line configure script

that I remember being on-site at a customer trying to debug at 2am

which turned out to be 'it expects exactly _1_ thing in usr/src with the letters 'zfs' in the filename)

and otherwise dies in a completely unscruitable fashion

oh no!

in it's way, it makes sd.c seem quaint and svelte :)

andyf: I've updated illumos/gcc #56 with responses to review comments; you may want to take a look at those and pull them into your illumos/gcc #57

be nice to get that in, thanks for working on it