Re: Armv7 linker failure on -current

On Sat, Jun 17, 2023 at 04:05:12PM -0700, Mark Millard wrote:
> On Jun 17, 2023, at 11:42, bob prohaska <[email protected]> wrote:
> 
> > On Sat, Jun 17, 2023 at 08:34:16AM -0700, Mark Millard wrote:
> >> On Jun 17, 2023, at 07:34, bob prohaska <[email protected]> wrote:
> >> 
> >>> A Pi2 running -current stops in buildworld with
> >> 
> >> I assume self building an update to main [so: 14].
> >> 
> > 
> > Correct.
> > 
> >> What commit last built successfully? (This bounds what would be different
> >> in official materials.)
> >> 
> > 
> > I haven't kept any manual records, but uname -a reports
> > FreeBSD www.zefox.com 14.0-CURRENT FreeBSD 14.0-CURRENT #39 main-9f135336f5: Fri Jun 16 02:18:02 PDT 2023     [email protected]:/usr/obj/usr/src/arm.armv7/sys/GENERIC arm
> > 
> > Since my original post git pull found a few updates and I restarted buildworld
> 
> I expect that this means there is no evidence left to look at
> what was originally reported.

I still have the buildworld output from the failure, but given what 
went before I'm not at all confident it represents anything to do
with FreeBSD. 

> 
> > It has failed also, but with what looks like an unrelated error. These problems
> > emerged after a silent hang using -j3, followed by power-cycling and reboot. -j3
> > was backed down to -j2 and buildworld restarted, with the mischief reported here
> > ensuing.
> 
> I see. Messy.
> 
> > I'm starting to wonder if there's some undetected corruption in either /usr/src
> > or /usr/obj. Perhaps I'd best run make cleandir twice and start over. That will
> > likely take at least  few days. This is a mechanical hard disk setup, with 1 GB
> > swap. Boot is via microSD using only bootcode.bin, with the usual msdos partition
> > on the hard disk.  
> > 
> > Is there a more expedient approach under the circumstances? 
> 
> If I understand right, there is no evidence to examine from
> the original report's context to guide any investigation. It

Apart from the buildworld log, no. I was quite surprised at the
lockup, -j3 and a gig of swap had been working fine for several
build/install cycles. It was especially surprising when nothing
turned up on the console.    

  

> seems you need to establish a known context (broken or
> working) to possibly investigate.
>

Make cleandir was run twice and -j2 meta-mode buildworld restarted.  
I'll be more careful with the wreckage next time.

Thanks for writing!

bob prohaska