Re: [Monotone-devel] mtn_automate usage

[Stephen Leake]

Timothy Brownawell <address@hidden> writes:

> On 07/14/2010 04:46 PM, Thomas Keller wrote:
>>
>> While I worked on nvm.validate-changes I had the idea to use
>> mtn_automate() in my pre-commit test hook to only scan the diffed source
>> part for CRLFs (see the other mail), but this usage is currently
>> prevented as we only allow mtn_automate() being called from
>> register_command().
>
> How bad would this actually be? Say you have your netsync notification
> hooks set to call 'heads' or some of the graph commands to print a
> pretty report of what was synced, and you also want a custom command
> that will commit and then sync...

+1

> Maybe instead of a strict "no recursion" we want to prevent recursion
> past some (arbitrary, large) depth like 100 or so, or keep a stack of
> all calls with options and arguments to check for cycles.

It would help to have a debug mode that just prints lua calls.

>   * things that affect the workspace, including writing out options.
>     everything other than writing out options tends to be user-explicit,
>     so we'd just need a check when writing options that we're the
>     outermost workspace-aware command. Possibly do this with something
>     similar to transaction_guard.

+1

-- 
-- Stephe