arsenal-devel team mailing list archive

[Bryce Harrington <bryce@xxxxxxxxxxxxx>]

On Tue, Dec 07, 2010 at 07:35:45AM -0800, Brian Murray wrote:
> On Tue, Dec 07, 2010 at 12:14:35AM -0800, Bryce Harrington wrote:
> > Bugs that are filed by or appended to by apport have key:value pairs
> > inserted into their descriptions.  The patch below enables arsenal to
> > parse this data out and provides it in a dictionary structure.
> 
> I was looking at a bug report the other day that was an apport-bug or
> apport-crash report whose description had been modified by the original
> reporter.  As a part of the modification they had somehow removed some
> of the line breaks so it looked something like:
> 
> Release: MaverickArchitecture: i386
> 
> You might want to think about how to handle that case.

If you can shoot me the bug # I can take a look and test this on it.

If the user did something like convert \n's to \r's, I can probably make
the split statement handle that type of situation.

If the newline was removed entirely though, that gets to be a pretty
tough problem, I'm not sure it can be feasibly be addressed other than
manually fixing up those types of bugs as they're spotted.

Note there's a few other things this code doesn't parse right yet.  For
example, it can't handle multiline values, like this:

Stacktrace:
 Using host libthread_db library "/lib/tls/i686/cmov/libthread_db.so.1".
 #0 0xb7689d5d in ?? ()
  
That just gets parsed as 'Stacktrace':'' so that's no good.

It also doesn't get the env info:

ProcEnviron:
 LANG=es_ES.UTF-8
 SHELL=/bin/bash

There's probably a few ways to handle these cases.  But frankly I don't
need to know the env, and prefer the stacktrace from the attachment, so
I can live without the data for now.  But would be nice to support.

Bryce