Re: [PATCH v13 00/23] Compile-time stack metadata validation

[Josh Poimboeuf <jpoimboe@redhat.com>]

Hi Ingo,

Do you still have objections to my proposed command-line interface for
stacktool?

As I described below, I still don't think subcommands are a good fit for
stacktool.  However, if you strongly disagree, I can change it.

The patch set is now quite mature at v14.  Further, it's holding up work
on several other items:

- more frame pointer fixes
- CFI generation and validation
- livepatch consistency model
- DWARF unwinder

So I'm wondering if it would be possible for you to merge stacktool in
time for the next merge window (4.5).

Thanks!
Josh

On Mon, Oct 12, 2015 at 09:23:14AM -0500, Josh Poimboeuf wrote:
> On Mon, Oct 12, 2015 at 09:41:11AM +0200, Ingo Molnar wrote:
> > 
> > * Josh Poimboeuf <jpoimboe@redhat.com> wrote:
> > 
> > > Hi Ingo,
> > > 
> > > Do you have any more objections to these patches?  Would you be willing
> > > to apply them?
> > 
> > So I still don't like the tool namespace you picked: Git-alike generic
> > naming plus subcommands work so much better that I'm not sure why we
> > are even having that discussion:
> 
> Because subcommands are useful in _some_ cases, but they aren't a
> panacea that should be blindly applied everywhere.
> 
> > if you name your tool 'stacktool' and
> > alias everything you have today to under 'stacktool run ...' and add
> > 'stacktool help' as a second, obvious subcommand then you'll have your
> > current syntax and a lot more future flexibility and ability to branch
> > off various functionality a'la Git, perf or kvmtool ...
> 
> Sure, subcommands work great for monolithic framework tools like git,
> perf, yum, docker, etc.  But stacktool is not (and never will be) a
> monolithic framework type of tool.
> 
> The suggestion to put 100% of the functionality under 'stacktool run
> [options]' is certainly possible.  But 'run' is so broad.  What else
> could the tool ever do but 'run'?