From: Ramkumar Ramachandra <artagnon@gmail.com>
Date: Wed, 03 Aug 2011 01:32:50 GMT
Subject: Re: [RFC/ PATCH] revert: Allow arbitrary sequencer instructions
Message-ID: <CALkWK0kfOjXeuswwOvgO5eotUE89rL-X4tTtf=frGR4WdR4iAA@mail.gmail.com>
In-Reply-To: <20110802205353.GF2743@elie>

Hi Jonathan,

Jonathan Nieder writes:
> Ramkumar Ramachandra wrote:
>> --- a/sequencer.h
>> +++ b/sequencer.h
>> @@ -32,6 +32,16 @@ struct replay_opts {
>>       size_t xopts_nr, xopts_alloc;
>>  };
>>
>> +struct replay_insn {
>> +     struct commit *commit;
>> +     enum replay_action action;
>> +};
>> +
>> +struct replay_insn_list {
>> +     struct replay_insn *item;
>> +     struct replay_insn_list *next;
>> +};
>
> Ah, so this allows sequences like
>
>        revert A
>        pick B
>        pick C
>        revert D

Exactly.

> Nit: why isn't the list-item struct something like
>
>        struct replay_insn item;
>        struct replay_insn_list *next;
>
> which would save a little memory management and memory access
> overhead (or even
>
>        enum replay_action action;
>        struct commit *operand;
>        struct replay_insn_list *next;
>
> since every "struct replay_insn" exists in the context of an
> insn list afaict)?

Right.  Good suggestion.

> Anyway, the general idea seems good.

Nice.  Thanks for the quick early feedback.  Much appreciated.

-- Ram