Re: Denial of denial of service

[prev] [thread] [next] [lurker] [Date index for 2007/01/28]

From: Yossi Kreinin
Subject: Re: Denial of denial of service
Date: 18:51 on 28 Jan 2007
Peter da Silva wrote:
> 
> If I run a command and I get "/some/obscure/file: No such file of 
> directory" I've got a place to start winkling at the problem. If I run a 
> command and I get "peter: Not a typewriter" I've got jack.
> 

One of my collegues started to work on a merge of pstack and strace, getting an 
strace printing call stacks from which the logged syscalls were made. This 
increases the chances of figuring out why the fuck /some/obscure/file is of any 
interest. When the executable code has symbols.

> I don't run a command and get a traceback, because nobody ever lets the 
> traceback go through to the end user, because it looks "unprofessional". 
> End users complain. They think the code ain't finished, because they 
> don't know that's one of the hateful things about software.
> 

I let the traceback (in managed environments) and core dumps (in unmanaged 
environments) go through to the end user, and the whole universe may 
professionally fuck itself.

End users usually like to know that their bug reports lead to bug fixes. All it 
takes is explaining them what people who won't expose them to such horrific 
things do with their bug reports.

I wish taking a snapshot of a process and debugging it were considered a feature 
worth implementing by those shipping programming environments 
(compilers/operating systems). One reason you don't get tracebacks when you run 
commands is that a lot of tools don't support that, not to mention state 
snapshots. Which wouldn't be rocket science in a whole lot of useful cases.
There's stuff above here

Generated at 23:01 on 06 Feb 2007 by mariachi 0.52