[darcs-devel] darcs patch: added postget to prefs (and 2 more)

[dagit at eecs.oregonstate.edu]

Quoting David Roundy <droundy at abridgegame.org>:

> No, working at a higher level won't address the postget "remote defaults"
> issue, what it will do is to make your run_posthook function unnecesary,
> which would mean that adding posthook support to a command will only
> require that the command add the posthook flag to its list of DarcsOptions,
> which seems like it would be nice, and the posthook would be guaranteed to
> always run on command success (i.e. if the command exits with ExitSuccess).
> It would mean that we'd have to be careful about when the darcs exits with
> ExitSuccess... i.e. the question of whether to exit successfully when there
> are no patches to pull/apply could become significant.

Could we get around this question by adding a command line switch to those
commands so that people can specify in defaults what those should return when
there are no patches?  We could default to exitting unsuccessfully when there
are no patches, and then let the user override this. I think exiting
unsuccessfully when there are no patches makes sense because when the repo
doesn't change, do you really need to run your script, right? But at the same
time, if you're logging darcs activity you may want to record anything and
everything.  On the other hand, returning the number of patches could be handy
too; the problem here is that we are no longer consistent with the rest of the
unix tools.  Another solution here is to always run the hook even when there
are no patches and then pass the number of patches to the script possibly in
the env.  The problem here is that we are offloading the complexity to the
scripts, but perhaps this is for the best since each user has a better idea how
they want to use darcs than we do.

> Actually, I'm thinking that if we implemented posthooks in this manner, we
> could avoid modifying the commands at all, and add the DarcsOption like we
> do the --help and --list-options flags in DarcsCommands itself.  Then every
> command would support --posthook, and they'd all behave in the same way.
>
> The trick, by the way, would be to use (in consider_running) code something
> like
>
>   do (command_command cmd) (FixFilePath fix_path:specops) extra
>            `Control.Exception.catch` foo specops
>      foo (ExitSuccess) specops
>   where foo (ExitException ExitSuccess) opts = run_posthook opts
>         foo e = Control.Exception.throwIO e
>
> which would mean that if the command either doesn't call exitWith (and just
> returns) or calls "exitWith ExitSuccess", the posthook will get run.

I've been keeping myself busy with other projects lately, but this is a good
idea, maybe I can find time to implement it over the weekend.

Thanks,
Jason