qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Eric Blake <eblake@redhat.com>
To: Eduardo Habkost <ehabkost@redhat.com>, qemu-devel@nongnu.org
Cc: Markus Armbruster <armbru@redhat.com>,
	Michael Roth <mdroth@linux.vnet.ibm.com>
Subject: Re: [Qemu-devel] [RFC 00/15] Error API: Flag errors in *errp even if errors are being ignored
Date: Tue, 27 Jun 2017 15:22:21 -0500	[thread overview]
Message-ID: <68640db9-3a9d-6542-3d1b-da62a1f11d9d@redhat.com> (raw)
In-Reply-To: <20170613165313.20954-1-ehabkost@redhat.com>

[-- Attachment #1: Type: text/plain, Size: 3221 bytes --]

On 06/13/2017 11:52 AM, Eduardo Habkost wrote:

> The Proposal
> ------------
> 
> I'm proposing replacing NULL errp with a special macro:
> IGNORE_ERRORS.  The macro will trigger special behavior in the
> error API that will make it not save any error information in the
> error pointer, but still keep track of boolean error state in
> *errp.
> 
> This will allow us to simplify the documented method to propagate errors
> from:
> 
>     Error *err = NULL;
>     foo(arg, &err);
>     if (err) {
>         handle the error...
>         error_propagate(errp, err);
>     }
> 
> to:
> 
>     foo(arg, errp);
>     if (ERR_IS_SET(errp)) {
>         handle the error...
>     }

Seems kind of neat to me!


> 
> This way, we can make ERR_IS_SET work even if errp was
> IGNORE_ERRORS.  The ERR_IS_* macros are reimplemented as:
> 
>   #define ERR_IS_SET(e) (*(e) && *(e) != &ignored_error_unset)
>   #define ERR_IS_IGNORED(e) (!(e) || *(e) == &ignored_error_unset || *(e) == &ignored_error_set)

Where the initial NULL checks go away if you get your way with a final
patch.

> 
> Ensuring errp is never NULL
> ---------------------------
> 
> The last patch on this series changes the (Error **errp)
> parameters in functions to (Error *errp[static 1]), just to help
> validate the existing code, as clang warns about NULL arguments
> on that case.  I don't think we should apply that patch, though,
> because the "[static 1]" syntax confuses Coccinelle.

Have you filed a bug report to the Coccinelle folks?  But yeah, it is
rather arcane C99 syntax that you don't see much of in common code.

> (Probably the easiest solution for that is to add assert(errp)
> lines to the ERR_IS_*() macros.)

Or even having the macros use a forced dereference through errp->xxx may
at least be enough for Coverity to catch things.

> Git branch
> ----------
> 
> This series depend on a few extra cleanups that I didn't submit
> to qemu-devel yet.  A git branch including this series is
> available at:
> 
>   git://github.com/ehabkost/qemu-hacks.git work/err-api-rework-ignore-ptr-v1
> 
> Eduardo Habkost (15):
>   tests: Test cases for error API
>   error: New IGNORE_ERRORS macro
>   Add qapi/error.h includes on files that will need it
>   [coccinelle] Use IGNORE_ERRORS instead of NULL as errp argument
>   qapi: Use IGNORE_ERRORS instead of NULL on generated code
>   test-qapi-util: Use IGNORE_ERRORS instead of NULL
>   Manual changes to use IGNORE_ERRORS instead of NULL
>   error: New ERR_IS_* macros for checking Error** values
>   [coccinelle] Use ERR_IS_* macros
>   test-qapi-util: Use ERR_IS_* macros
>   Manual changes to use ERR_IS_* macros
>   error: Make IGNORED_ERRORS not a NULL pointer
>   rdma: Simplify var declaration to avoid confusing Coccinelle
>   [coccinelle] Eliminate unnecessary local_err/error_propagate() usage
>   [test only] Use 'Error *err[static 1]' instead of 'Error **errp' to
>     catch NULL errp arguments

I have not reviewed the series yet, but the idea looks promising.

-- 
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3266
Virtualization:  qemu.org | libvirt.org


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 604 bytes --]

  parent reply	other threads:[~2017-06-27 20:22 UTC|newest]

Thread overview: 41+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-06-13 16:52 [Qemu-devel] [RFC 00/15] Error API: Flag errors in *errp even if errors are being ignored Eduardo Habkost
2017-06-13 16:52 ` [Qemu-devel] [RFC 01/15] tests: Test cases for error API Eduardo Habkost
2017-06-13 16:53 ` [Qemu-devel] [RFC 02/15] error: New IGNORE_ERRORS macro Eduardo Habkost
2017-06-13 16:53 ` [Qemu-devel] [RFC 03/15] Add qapi/error.h includes on files that will need it Eduardo Habkost
2017-06-13 16:53 ` [Qemu-devel] [RFC 04/15] [coccinelle] Use IGNORE_ERRORS instead of NULL as errp argument Eduardo Habkost
2017-06-13 16:53 ` [Qemu-devel] [RFC 05/15] qapi: Use IGNORE_ERRORS instead of NULL on generated code Eduardo Habkost
2017-06-13 16:53 ` [Qemu-devel] [RFC 06/15] test-qapi-util: Use IGNORE_ERRORS instead of NULL Eduardo Habkost
2017-06-13 16:53 ` [Qemu-devel] [RFC 07/15] Manual changes to use " Eduardo Habkost
2017-06-13 16:53 ` [Qemu-devel] [RFC 08/15] error: New ERR_IS_* macros for checking Error** values Eduardo Habkost
2017-06-13 16:53 ` [Qemu-devel] [RFC 09/15] [coccinelle] Use ERR_IS_* macros Eduardo Habkost
2017-06-13 16:53 ` [Qemu-devel] [RFC 10/15] test-qapi-util: " Eduardo Habkost
2017-06-13 16:53 ` [Qemu-devel] [RFC 11/15] Manual changes to use " Eduardo Habkost
2017-06-13 16:53 ` [Qemu-devel] [RFC 12/15] error: Make IGNORED_ERRORS not a NULL pointer Eduardo Habkost
2017-06-13 16:53 ` [Qemu-devel] [RFC 13/15] rdma: Simplify var declaration to avoid confusing Coccinelle Eduardo Habkost
2017-06-13 16:53 ` [Qemu-devel] [RFC 14/15] [coccinelle] Eliminate unnecessary local_err/error_propagate() usage Eduardo Habkost
2017-06-13 16:53 ` [Qemu-devel] [RFC 15/15] [test only] Use 'Error *err[static 1]' instead of 'Error **errp' to catch NULL errp arguments Eduardo Habkost
     [not found]   ` <20170615121407.GA2399@work-vm>
2017-06-17 19:33     ` Eduardo Habkost
2017-06-19  8:48       ` Dr. David Alan Gilbert
2017-06-19  9:43         ` Peter Maydell
2017-06-19 13:26           ` Eduardo Habkost
2017-06-27 20:12             ` Eric Blake
2017-06-27 21:31               ` Eduardo Habkost
2017-06-27 20:22 ` Eric Blake [this message]
2017-06-28  9:05 ` [Qemu-devel] [RFC 00/15] Error API: Flag errors in *errp even if errors are being ignored Markus Armbruster
2017-06-28 17:41   ` Eduardo Habkost
2017-06-29  6:54     ` Markus Armbruster
2017-06-29 12:57       ` Eduardo Habkost
2017-06-30 11:40         ` Markus Armbruster
2017-07-01 14:20           ` Eduardo Habkost
2017-07-03 12:51             ` Markus Armbruster
2017-07-01 14:29           ` Eduardo Habkost
2017-07-03 13:21             ` Markus Armbruster
2017-07-03 13:47               ` Eduardo Habkost
2017-06-29 14:01     ` Daniel P. Berrange
2017-06-29 13:39   ` Paolo Bonzini
2017-06-29 14:18     ` Daniel P. Berrange
2017-06-29 17:09       ` Eduardo Habkost
2017-06-29 17:38         ` Daniel P. Berrange
2017-06-29 17:47           ` Eduardo Habkost
2017-06-29 18:04             ` Daniel P. Berrange
2017-06-29 14:14   ` Daniel P. Berrange

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=68640db9-3a9d-6542-3d1b-da62a1f11d9d@redhat.com \
    --to=eblake@redhat.com \
    --cc=armbru@redhat.com \
    --cc=ehabkost@redhat.com \
    --cc=mdroth@linux.vnet.ibm.com \
    --cc=qemu-devel@nongnu.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 NNTP newsgroup(s).