portsdb and ruby bug on 4-STABLE

[Vulpes Velox]

On Mon, 6 Sep 2004 16:24:57 +0300 (EEST)
Dmitry Pryanishnikov <dmitry at atlantis.dp.ua> wrote:

> 
> Hello!
> 
>   I'm hitting notoriuos bug during the portsdb -u last several days:
> 
> root at core# portsdb -fu
> [Updating the portsdb <format:bdb1_btree> in /usr/ports ... - 10078
> port entries found 
> .........1000.........2000.........3000.........4000.........5000..
> .......6000.........7000../usr/local/lib/ruby/site_ruby/1.8/portsdb
> .rb:587: [BUG] Segmentation fault
> ruby 1.8.2 (2004-07-29) [i386-freebsd4]
> Abort trap (core dumped)
> 
> Machine has ECC memory and is rock-solid, so I'd say it's definitely
> a software bug. Machine is running 4.9-RELEASE-p11, the same
> behaviour is seen on another PC, running 4.10-RELEASE-p2. However,
> my third server, running 4.7-RELEASE, doesn't trap with the same
> ports collection and the same INDEX generation procedure (I use
> portindex). Versions of the involved software are the same on all
> three machines:
> 
> portindex-18_1      Incremental ports INDEX file builder
> portupgrade-20040701_3 FreeBSD ports/packages administration and
> management python-2.3.4_2      An interpreted object-oriented
> programming language ruby-1.8.2.p2_1     An object-oriented
> interpreted scripting language ruby18-bdb1-0.2.2   Ruby interface to
> Berkeley DB revision 1.8x with full
> 
> So I'm curious, why this bug doesn't show under 4.7-RELEASE but
> shows under 4.9+? libc issue?
> 
> P.S. BTW ruby sometimes runs into infinite loop instead of crashing
> (if it helps someone). Of course, I use portindexdb as a workaround.

fix it using rm INDEX* and then rebuild the index using portindex and
then rebuild the INDEX.db using portindexdb. Port upgrade now works
again. I have not seen this problem on for X, but did once on
5.3beta3.