Re: [Qemu-devel] [PATCH v3] qemu-io: Add generic function for reinitializing optind.

[Eric Blake <eblake@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 1758 bytes --]

On 1/9/19 7:44 AM, Daniel P. Berrangé wrote:
> On Wed, Jan 09, 2019 at 01:23:01PM +0000, Richard W.M. Jones wrote:
>> How about this one?  Add a generic osdep function for reinitializing
>> optind, which does optreset on FreeBSD (but is identical on all other
>> OSes).  Use it from qemu-io and qemu-img.
>>
>> I have tested this on Linux, FreeBSD and OpenBSD.
>>

> 
>> WARNING: architecture specific defines should be avoided
>> #78: FILE: include/qemu/osdep.h:600:
>> +#ifdef __FreeBSD__
> 
> Normally we'd suggest doing a configure test to for the platform
> feature and then using a feature based ifdef test. In this case
> though that would be difficult and/or overly complex.
> 
> This does make me wonder about the other *BSDs, OS-X and Mingw
> though ?  Should they all be using the #else codepath, or should
> the other BSDs / OS-X use the __FreeBSD__ codepath.

Indeed, and I already suggested a configure-time probe  on the v2
review.  My preference, if we want to go with this helper function for a
hard reset, would be:

#if HAVE_OPTRESET
  optind = 1;
  optreset = 1;
#else
  optind = 0;
#endif

where HAVE_OPTRESET is based on a configure-time probe.

Basically, any platform that HAS optreset will probably honor it; any
platform that lacks optreset is hopefully okay with the
POSIX-unspecified behavior of optind = 0.

But I do like the idea presented in this version of trying to isolate
the reset into a common helper function, so that if we have to make
later changes, we only have to touch the one function rather than all
callsites that reset getopt parsing.

-- 
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3226
Virtualization:  qemu.org | libvirt.org


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]