[minicoredumper] Debugging corrupt tar files (invalid sparse archive member)

[John Ogness]

On 2023-02-09, John Ogness <john.ogness at linutronix.de> wrote:
> However, hearing your argument about not being able to use gzip, I am
> wondering if falling back to avoiding tar is the correct
> workaround. In your case, you are expecting the dumps to be quick and
> if it suddenly falls back to just gzip, you will suddenly have minutes
> of downtime instead of seconds.

So I implemented an alternative workaround that truncates the core file
instead of falling back to gzip. The changes to the code are also quite
small, so I am comfortable using this workaround instead.

However, in a core file, most of the interesting data is at the _end_ of
the core file. This is the part that is truncated. So all you end up
with is 8GB of heap data and the line of code where the crash
occurred. But you may have no stack data available. (I suppose if you
are allocating thread stacks on the heap, they will probably be
available as well.)

It is better than nothing, which is what you currently end up with.

For users that want the full core file, they can disable in_tar support
for that particular application. For 2.1 I will add support for a more
modern tar format.

Unless there are any objections, this is how I will workaround this for
the 2.0.3 bugfix release.

John Ogness