Re: [PATCH 0/2] addr2line: new option -n to add a newline at the end

[Maurizio Papini <mpapini@redhat.com>]

Hi Nick and thanks for your time.

> I am sorry - I totally missed this patch submission. :-(

The patch is definitely not urgent, trivial stuff indeed, and my submission day
was so close to the end of 2022.

> I am not sure what you are getting at here.  Are you saying that addr2line
> can get stuck and so a timeout is needed in order to be to distinguish between
> addr2line being slow and addr2line being stuck ?  If so then this sounds
> like a more serious problem that needs to be addressed by something other
> than just adding a blank line to the output.

No, I didn't see this kind of problem. I'm talking about this use case:
addr2line as a process, with "-i" option, that is reading from stdin
working on a
big binary, say like the Linux kernel with debug info (~400M+), so that spawning
the process once is definitely better than doing it every time.

> I have to say that I am kind of with Jan on this one - it seems like
> a lot of effort to fix a small problem.  Isn't there an easier way
> that would not involve patching addr2line ?  For example if you use
> the -a/--addresses option then each decoded address will be prefixed
> with a line containing a simple hex number.  This would allow a consumer
> of addr2line's output to detect the start of each decoded address.

I'm thinking about using your suggestion (thanks): with -a the address is added
as a prefix as you said, while it would have been useful to have it at
the end :-),
but querying two times...

Maurizio

Maurizio


On Tue, Jan 10, 2023 at 1:19 PM Nick Clifton <nickc@redhat.com> wrote:
>
> Hi Maurizio,
>
> > Do you think this should go through an RFC? Do you have any thoughts?
>
> I am sorry - I totally missed this patch submission. :-(
>
> >> This series adds a new option to addr2line (-n) to append a newline
> >> after the last informative one.
>
> I have to say that I am kind of with Jan on this one - it seems like
> a lot of effort to fix a small problem.  Isn't there an easier way
> that would not involve patching addr2line ?  For example if you use
> the -a/--addresses option then each decoded address will be prefixed
> with a line containing a simple hex number.  This would allow a consumer
> of addr2line's output to detect the start of each decoded address.
>
>
> > the
> > additional empty line can be used to mark the end of the inlined functions
> > lists so that an application can get the output without defining a timeout.
>
> I am not sure what you are getting at here.  Are you saying that addr2line
> can get stuck and so a timeout is needed in order to be to distinguish between
> addr2line being slow and addr2line being stuck ?  If so then this sounds
> like a more serious problem that needs to be addressed by something other
> than just adding a blank line to the output.
>
> Cheers
>    Nick
>
>