From: Anthony Green <green@moxielogic.com>
To: Andrew Haley <aph@redhat.com>
Cc: libffi-discuss <libffi-discuss@sourceware.org>
Subject: Re: How to call 'printf' using libffi?
Date: Sat, 20 Mar 2021 19:56:57 -0400 [thread overview]
Message-ID: <CACxje59Ost3P-zm5HequWmxi9UNXmavLgV5Ru+JcVuZPVDY5SQ@mail.gmail.com> (raw)
In-Reply-To: <CACxje59g9JirgoTxohQOHrP+At5L-JypHmyJZ7GTtd6cCmSqRA@mail.gmail.com>
It's worth noting that JNA, one of libffi's big users, does the
float-to-double promotion at their layer.
https://github.com/java-native-access/jna/issues/463
Ruby's FFI does the same thing:
https://github.com/ffi/ffi/blob/6d14c0a9107c0d5febb3bf92a60d2581e768fa2f/ext/ffi_c/Variadic.c
This is making me feel more comfortable about returning an error on float
varargs. (I've change my mind about aborting)
AG
On Sat, Mar 20, 2021 at 6:40 PM Anthony Green <green@moxielogic.com> wrote:
> It turns out there are problems with ffi_prep_cif_var(). One issue is
> that libffi doesn't promote floats to doubles for varargs, like most C
> hackers expect. There was a bug filed against that recently:
> https://github.com/libffi/libffi/issues/608
> Should libffi do the promotion, or leave it as an exercise to the
> developer? Right now I'm thinking that it should simply abort in
> ffi_prep_cif_var() if passed any of ffi_type_float, ffi_type_*short, or
> ffi_type_*char and require that the caller handle promotions and only pass
> pointers, structs, doubles, ints or larger, as it seems that we're not
> promoting short integer values either. If I pass a very long list of
> varargs that end in a sequence of ffi_type_schar, for instance, the callee
> doesn't find the right values in the right stack locations. So maybe we
> shouldn't allow that in the first place.
>
> I'm interested in opinions.
>
> Thanks,
>
> AG
>
>
> On Fri, Mar 19, 2021 at 7:43 AM Andrew Haley <aph@redhat.com> wrote:
>
>> On 3/19/21 11:01 AM, Anthony Green wrote:
>> >>> ffi_prep_cif_var(&cif, FFI_DEFAULT_ABI, 1, 3, &ffi_type_sint,
>> >> arg_types);
>> >>
>> >> Surely that only works if a varargs call uses the same ABI as a normal
>> >> call. They don't always,
>> >>
>> > I'm using the '_var' variant of ffi_prep_cif that takes this into
>> account.
>>
>> Aha! Missed it.
>>
>> --
>> Andrew Haley (he/him)
>> Java Platform Lead Engineer
>> Red Hat UK Ltd. <https://www.redhat.com>
>> https://keybase.io/andrewhaley
>> EAC8 43EB D3EF DB98 CC77 2FAD A5CD 6035 332F A671
>>
>>
next prev parent reply other threads:[~2021-03-20 23:57 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-03-18 14:04 ShaJunxing
2021-03-18 15:25 ` Anthony Green
2021-03-19 2:08 ` ShaJunxing
2021-03-19 2:36 ` Kaz Kylheku (libffi)
2021-03-19 3:09 ` ShaJunxing
2021-03-19 10:13 ` Andrew Haley
2021-03-19 11:01 ` Anthony Green
2021-03-19 11:43 ` Andrew Haley
2021-03-20 22:40 ` Anthony Green
2021-03-20 23:56 ` Anthony Green [this message]
2021-03-22 10:36 ` Andrew Haley
2021-03-19 11:26 ` Jarkko Hietaniemi
2021-03-19 11:33 ` Anthony Green
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=CACxje59Ost3P-zm5HequWmxi9UNXmavLgV5Ru+JcVuZPVDY5SQ@mail.gmail.com \
--to=green@moxielogic.com \
--cc=aph@redhat.com \
--cc=libffi-discuss@sourceware.org \
/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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).