From: Patrick Steinhardt <ps@pks.im>
To: The development of GNU GRUB <grub-devel@gnu.org>
Cc: Max Tottenham <mtottenh@akamai.com>, Daniel Kiper <dkiper@net-space.pl>
Subject: Re: [PATCH v2 2/6] json: Implement wrapping interface
Date: Thu, 7 Nov 2019 07:26:43 +0100 [thread overview]
Message-ID: <20191107062643.GA2544@ncase> (raw)
In-Reply-To: <20191107025130.GA64831@lon-mpk47.lan>
[-- Attachment #1: Type: text/plain, Size: 2707 bytes --]
On Thu, Nov 07, 2019 at 02:51:30AM +0000, Max Tottenham via Grub-devel wrote:
> On 11/06, Patrick Steinhardt wrote:
> > On Wed, Nov 06, 2019 at 02:57:52PM +0000, Max Tottenham wrote:
> > > I had one more comment about this patch which I forgot to send
> > > last time:
> > >
> > > +struct grub_json
> > > +{
> > > + void *tokens;
> > > + char *string;
> > > + grub_size_t idx;
> > > +};
> > > +typedef struct grub_json grub_json_t;
> > >
> > >
> > > Why is tokens declared as void* rather than jsmntok_t* ? There are a bunch of
> > > casts back to jsmntok_t * as part of your wrapper.
> > >
> > > Is the idea to attempt to provide a separation between grub_json_* and jsmn (to
> > > allow other libraries to implement the JSON module?). Or is there some sort of
> > > pointer arithmetic I'm missing that requires the use of void*?
> > >
> > > If it's the former, why don't we just do:
> > >
> > > typedef jsmntok_t grub_json_tok_t;
> > > struct grub_json
> > > {
> > > grub_json_tok_t *tokens;
> > > char *string;
> > > grub_size_t idx;
> > > };
> > > typedef struct grub_json grub_json_t;
> > >
> > >
> > > For now and worry about other libraries later?
> >
> > The reason is that we'd have to include "jsmn.h" in
> > "include/grub/json.h" to make `jsmntok_t` available. I definitely
> > want to avoid this in order to not leak any decls from jsmn into
> > users of "json.h" and to be able to have "jsmn.h" live in
> > "grub-core/lib/json". It's also not possible to have a forward
> > declaration here due to how `jsmntok_t` is declared.
> >
> > Patrick
>
>
> In which case, looking again at this patchset - grub_json_t
> is always used as an opaque value by consumers of json.h .
>
> Why not put a forward declaration of grub_json_t in include/grub/json.h,
> and stick the implementation in grub-core/lib/json.c :
>
> include/grub/json.h:
>
> typedef struct grub_json_t grub_json_t;
>
> grub-core/lib/json.c:
> ...
> #include <grub/json.h>
> #include "jsmn.h"
> ...
> struct grub_json_t {
> jsmntok_t* tokens;
> ...
> };
>
>
> Unless there is something I'm missing doesn't the above achieve the
> desired result?
If we'd do that, all calls to `grub_json_get{child,value}` would
have to return a dynamically allocated `grub_json_t` structure as
the storage size of `grub_json_t` would not be known at compile
time and thus cannot be allocated on the caller's stack. Right
now, the caller can just declare the structure on-stack and have
the interface fill up these structures for him, without any need
for dynamic memory allocation.
Patrick
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]
next prev parent reply other threads:[~2019-11-07 6:26 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-11-05 12:54 [PATCH v2 2/6] json: Implement wrapping interface Max Tottenham
2019-11-05 13:12 ` Patrick Steinhardt
2019-11-06 11:53 ` Daniel Kiper
2019-11-06 14:57 ` Max Tottenham
2019-11-06 20:20 ` Patrick Steinhardt
2019-11-07 2:51 ` Max Tottenham
2019-11-07 6:26 ` Patrick Steinhardt [this message]
2019-11-08 13:30 ` Max Tottenham
2019-11-10 16:44 ` Patrick Steinhardt
2019-11-13 11:43 ` Daniel Kiper
-- strict thread matches above, loose matches on Subject: below --
2019-11-02 18:06 [PATCH 0/6] Support for LUKS2 disc encryption Patrick Steinhardt
2019-11-05 6:58 ` [PATCH v2 0/6] Support for LUKS2 disk encryption Patrick Steinhardt
2019-11-05 6:58 ` [PATCH v2 2/6] json: Implement wrapping interface Patrick Steinhardt
2019-11-05 9:54 ` Max Tottenham
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20191107062643.GA2544@ncase \
--to=ps@pks.im \
--cc=dkiper@net-space.pl \
--cc=grub-devel@gnu.org \
--cc=mtottenh@akamai.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.