Jonathan Tan <jonathantanmy@xxxxxxxxxx> writes:
> "Glen Choo via GitGitGadget" <gitgitgadget@xxxxxxxxx> writes:
>> +static inline void config_reader_push_source(struct config_source *top)
>> +{
>> + if (cf_global)
>> + top->prev = cf_global;
>
> Don't we want to set prev unconditionally here (i.e. set it to NULL if
> cf_global was NULL)?
>
>> + cf_global = top;
>> +}
>> +
>> +static inline struct config_source *config_reader_pop_source()
>> +{
>> + struct config_source *ret;
>> + if (!cf_global)
>> + BUG("tried to pop config source, but we weren't reading config");
>> + ret = cf_global;
>> + cf_global = cf_global->prev;
>> + return ret;
>> +}
>
> ...since we use it unconditionally here.
Ah, thanks. It's safe as of this patch, since we always zero-initialize
"struct config_source" and nothing else sets ".prev", but this asymmetry
is a bug waiting to happen.
