should loader be re-installed on the both mirrored drives in pool, in order to boot from them, if prtvtoc | fmthard and zpool attach , have been already done (both drives detached and resivlered)

btw, as people work on LP64'izing, it looks like illumos still needs to fix instances of the bug originally reported as defect.opensolaris.org/bz/show_bug.cgi?id=18153 in which calls of execl(...., 0) are made that pass a 32-bit int where the function is looking for a 64-bit NULL pointer

[illumos-gate] 14993 zfs-tests: Move platform independent tests to a shared runfile -- Ryan Moeller <ryan⊙ic>

alanc: wait, Solaris wasn't LP64 from the get go? I swear everything I've read says it was

KungFuJesus: Nope.

what was it? ILP64?

ILP32

Well, technically both, right? 32/64.

Well, originally it was ILP32, then LP64 execution environments were added.

Most commands in illuos are still ILP32 while at some point in the S11 train they are all LP64.

Yes, Solaris was exclusively ILP32 until Solaris 7 started adding LP64 stuff

blogs.oracle.com/solaris/post/movin…g-oracle-solaris-to-lp64-bit-by-bit is the backstory

blogs.oracle.com/solaris/post/oracl…14-beta-progress-on-lp64-conversion has the chart showing our big change in 11.4 (which was orignally going to be 12.0 when we did most of that work)

There was a brief ILP64 offshoot of Solaris made by HAL for their SPARC64 processors before Sun did the LP64 in Solaris itself - queue.acm.org/detail.cfm?id=1165766 has more on that story

there is so much 64bit uncleanliness

so much

if someone wanted directing toward some, I happen to have a list

a lot of it is probably people passing cookies in (void *)'s and casting them straight to int.

so there's even a lot of bitesize stuff in there

Is there a smatch/gcc warning for those by chance?

not when 32bit

when 64bit gcc of course gives pointer cast to different sized int

pmooney: having -fforce-omit-frame-pointer seems reasonable for your use-case

jperkin: cool, thanks

I do appreciate -fomit-frame-pointer being a lie for things like pkgsrc, where it would be a huge pain to teach all those packages to skip it

but for the stuff which truly needs it...

it is a painful way to discover something uses the base pointer as scratch