Christian Kellermann wrote:

> PS: I just found out that when you do a darcs send --edit and quit
> vim (in my case) with :q! the sending is not aborted...

I've seen this bite a few people on the list lately, resulting in
accidental (and potentially embarrassing) sends.

Should darcs send with an explicit --edit abort if the description is
not edited (i.e. its checksum hasn't changed after $VISUAL exits)?

Is there some other way we can address this workflow/usability issue?

On Wed, Oct 29, 2008 at 6:27 PM, Trent Buck <bugs@darcs.net> wrote:
>
> New submission from Trent Buck <trentbuck@gmail.com>:
>
> Christian Kellermann wrote:
>
>> PS: I just found out that when you do a darcs send --edit and quit
>> vim (in my case) with :q! the sending is not aborted...
>
> I've seen this bite a few people on the list lately, resulting in
> accidental (and potentially embarrassing) sends.
>
> Should darcs send with an explicit --edit abort if the description is
> not edited (i.e. its checksum hasn't changed after $VISUAL exits)?
>
> Is there some other way we can address this workflow/usability issue?

I thought there was a proposal floating around in the bug tracker that
darcs should ask one last time after the editor exits.  "Okay, about
to send potentially embarrassing patches.  Take a 5 minute break and
then answer this question. Are you really, really, sure that you spell
checked and don't need to amend these patches?"

Even better, maybe we could have darcs ask you a series of questions...

Before you send these patches, what is 5 x 7? 35
Okay, you seem to be awake, but what is the capital of Singapore? Singapore City
Alright, send patch bundle now? y
Sending patches

Okay that's just silly, but I know gmail has a plugin like this, it
makes you do some math problems if you're sending email at odd hours
:)

But yeah, adding a simple yes/no question at the end with a default of
no would be nice.  I know I've wanted it on more than one occasion.

This should be very easy to fix, but may require tweaking some of the
tests.  This would be a good aspiring darcs hacker task.

Jason

On Wed, Oct 29, 2008 at 06:47:26PM -0700, Jason Dagit wrote:
> On Wed, Oct 29, 2008 at 6:27 PM, Trent Buck <bugs@darcs.net> wrote:
> >
> > New submission from Trent Buck <trentbuck@gmail.com>:
> >
> > Christian Kellermann wrote:
> >
> >> PS: I just found out that when you do a darcs send --edit and quit
> >> vim (in my case) with :q! the sending is not aborted...
> >
> > I've seen this bite a few people on the list lately, resulting in
> > accidental (and potentially embarrassing) sends.
> >
> > Should darcs send with an explicit --edit abort if the description is
> > not edited (i.e. its checksum hasn't changed after $VISUAL exits)?
> >
> > Is there some other way we can address this workflow/usability issue?
> 
> I thought there was a proposal floating around in the bug tracker that
> darcs should ask one last time after the editor exits.  "Okay, about
> to send potentially embarrassing patches.  Take a 5 minute break and
> then answer this question. Are you really, really, sure that you spell
> checked and don't need to amend these patches?"

What I'm proposing is that we refine the "it looks like the user
screwed up" heuristic to include "asked to --edit-description, but
didn't."

We could also change it so that instead of aborting when dumb-user-p
is true, we prompt:

    $ darcs send -a                    # sends without prompting
    $ VISUAL=cat darcs send -a --edit  # matches dumb-user-p

    Darcs send was called with --edit-description, but
    you didn't edit the description!

    Continue? [y/N]

I'm against having darcs *always* ask, because then I'll just get in
the habit of saying "yes" without thinking... meaning that the
question just becomes another piece of meaningless interaction.

On Thu, Oct 30, 2008 at 01:27:16 -0000, Trent Buck wrote:
> > PS: I just found out that when you do a darcs send --edit and quit
> > vim (in my case) with :q! the sending is not aborted...

I realise this does not solve the issue at hand, but as a user, I have
trained myself to killall darcs whenever I want to abort a send.
Quitting the text editor is then rendered harmless.

(The book Producing Open Source Software mentions that users will tend
to find workarounds to bugs and live with them unhappily rather than
complain about them, and that we should watch out for that... so this is
an example of me being such a user)

I've been doing the same. Definitely an issue that people run into.

On Thu, Oct 30, 2008 at 09:05:43AM -0000, Eric Kow wrote:
> On Thu, Oct 30, 2008 at 01:27:16 -0000, Trent Buck wrote:
> > > PS: I just found out that when you do a darcs send --edit and quit
> > > vim (in my case) with :q! the sending is not aborted...
> 
> I realise this does not solve the issue at hand, but as a user, I have
> trained myself to killall darcs whenever I want to abort a send.
> Quitting the text editor is then rendered harmless.

Same for me.

The following patch updated the status of issue1186 to be resolved:

* Resolve issue1186: give a chance to abort if user did not edit description file. 
Ignore-this: 84a1431b376c91e135d68094dee25e44

When user does not change file for send or record description ask if
darcs should continue.