Bug 104090 – herdstat consumes all cpu/memory

• Product: Gentoo Linux
• Status: RESOLVED FIXED

[Description Rob Cakebread (RETIRED) 2005-08-28 20:44:11 UTC]

Heres a fun one for you:
herdstat -q -n -m dev-db/postgis

and your cpu will go to 99% and memory slowly be used completely.
Careful if you have 128meg or less, it'll use all your memory quickly.
I let it run several minutes on a 1 gig machine but it never finished, consuming
80% memory.

I tried it on two machines with 1.1.1_p3 1.1.1_p5
Its only affected by dev-db/postgis since the package was updated today

[Comment 1 Aaron Walker (RETIRED) 2005-08-29 04:19:16 UTC]

ewww.  that was the last set of command-line options I was expecting to see when
reading the summary (since it's not at all cpu/mem intensive like the -p option).  

I am able to reproduce it here though, so that's good.  Just got home from work,
so I'll have a closer look in a bit.

[Comment 2 Aaron Walker (RETIRED) 2005-08-30 06:04:48 UTC]

Here's our culprit (line 36 of postgis-1.0.1_p20050805.ebuild):

xslv="$(echo ${xslv} | cut -d'-' -f1 )-$(echo ${xslv} | cut -d'-' -f2 )-$(echo
${xslv} | cut -d'-' -f3)"

herdstat tries to do /very/ simple variable substitution so that stuff like

HOMEPAGE="...${PV}..."

etc, get substituted and displayed to the user correctly.  For some reason, in
this particular case, the recursion depth safeguard I had in place didn't do
what is was supposed to (break out of the loop if it cannot resolve the variable
after recursing a certain number of times), thus the inf. loop.

I'll do a 1.1.1_p6 some time today.

[Comment 3 Aaron Walker (RETIRED) 2005-08-30 19:00:28 UTC]

1.1.1_p6 in cvs.

[Comment 4 Jakub Moc (RETIRED) 2006-09-24 02:51:55 UTC]

Assuming fixed, closing.