[PATCH] Introduce new filtering architecture

[Nikolay Marchuk]

On 20.06.2017 08:42, Dmitry V. Levin wrote:
> On Mon, Jun 19, 2017 at 08:37:32PM +0700, Nikolay Marchuk wrote:
>> On 19.06.2017 11:08, Eugene Syromiatnikov wrote:
>>>> +		if (!strncmp(name, str, len)) {
>>> Not sure whether strncmp is better than strcmp here, since this function
>>> is not called with user-supplied str at all.
>>>
>>> Or, do you expect that these will work with user-supplied strings in the
>>> future?
>>>
>>> BTW, personally, I don't think that this needs to be string-keyed.
>>> Something as simple as
>>>
>>>     enum filter_type_name {
>>>             FILTER_TYPE_syscall,
>>>             FILTER_TYPE_fd,
>>>     };
>>>
>>>     #define FILTER_TYPE(name) \
>>>             [FILTER_TYPE_ ## name] = { \
>>>                     .parse_filter = parse_ ## name ## _filter, \
>>>                     .run_filter = run_ ## name ## _filter, \
>>>                     ._free_priv_data = free_ ## name ## _filter, \
>>>             }
>>>
>>>     static const struct filter_type {
>>>             void* (*parse_filter)(const char *, const char *const);
>>>             bool (*run_filter)(struct tcb *, void *);
>>>             void (*_free_priv_data)(void *);
>>>     } filter_types[] = {
>>>             FILTER_TYPE(syscall),
>>>             FILTER_TYPE(fd),
>>>     };
>>>
>>>     struct filter *
>>>     add_filter_to_array(struct filter **filters, unsigned int *nfilters,
>>>                         enum filter_type_name name)
>>>     {
>>>             *filters = xreallocarray(*filters, ++(*nfilters),
>>>                                      sizeof(struct filter));
>>>             struct filter *filter = &((*filters)[*nfilters - 1]);
>>>             filter->type = filters[name];
>>>             return filter;
>>>     }
>>>
>>>     [...]
>>>
>>>     #define create_filter(a, t) actual_create_filter(a, FILTER_TYPE_ ## t)
>>>
>>>     [...]
>>>
>>>     create_filter(action, fd);
>>>
>>> should also suffice.
>> I am going to use string-keyed filters and actions for new filtering
>> language parsing and enum is not sufficient.
> 
> I'm not sure I follow this; how could it be that enum is not sufficient?
This functions will be later used for parsing of user-supplied expression.
If enum is used, I will need additional table to parse string to enum,
because preprocessor wrapper will not work for user-supplied strings.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 213 bytes
Desc: OpenPGP digital signature
URL: <http://lists.strace.io/pipermail/strace-devel/attachments/20170620/ee9da991/attachment.bin>