DPDK patches and discussions
 help / color / mirror / Atom feed
From: "Morten Brørup" <mb@smartsharesystems.com>
To: "Konstantin Ananyev" <konstantin.ananyev@huawei.com>,
	"Thomas Monjalon" <thomas@monjalon.net>
Cc: <dev@dpdk.org>, <ferruh.yigit@amd.com>,
	<stephen@networkplumber.org>, <bruce.richardson@intel.com>,
	<david.marchand@redhat.com>
Subject: RE: Coding Style for local variables
Date: Thu, 20 Jun 2024 11:02:21 +0200	[thread overview]
Message-ID: <98CBD80474FA8B44BF855DF32C47DC35E9F540@smartserver.smartshare.dk> (raw)
In-Reply-To: <978413ec22cb4b76872bd85f3588e5f3@huawei.com>

> From: Konstantin Ananyev [mailto:konstantin.ananyev@huawei.com]
> 
> > > From: Thomas Monjalon [mailto:thomas@monjalon.net]
> > >
> > > 10/06/2024 18:31, Konstantin Ananyev:
> > > > Morten said:
> > > > > The coding style guide says:
> > > > >
> > > > > "Variables should be declared at the start of a block of code rather
> than
> > > in the middle. The exception to this is when the variable is
> > > > > const in which case the declaration must be at the point of first
> > > use/assignment. Declaring variable inside a for loop is OK."
> > > > >
> > > > > Since DPDK switched to C11, variables can be declared where they are
> used,
> > > which reduces the risk of using effectively uninitialized
> > > > > variables. "Effectively uninitialized" means initialized to 0 or NULL
> > > where declared, to silence any compiler warnings about the use of
> > > > > uninitialized variables.
> > > > >
> > > > > Can we please agree to remove the recommendation/requirement to
> declare
> > > variables at the start of a block of code?
> > > >
> > > > I know that modern C standards allow to define variable in the middle.
> > > > But I am strongly opposed to allow that in DPDK coding style.
> > > > Such practice makes code much harder to read and understand (at least
> for
> > > me).
> > >
> > > Yes it is convenient to know that all variables are described
> > > in a known place, just after function parameters.
> > >
> > > There is also a consistency concern.
> > >
> > > Old contributors like to be in a comfort zone,
> > > 	and we don't want to lose old contributors.
> > > New contributors may be refrained by old rules,
> > > 	and we would like to get more new contributors.
> > >
> > > So that's a tricky decision.
> > >
> >
> > Independent research shows that readability is improved by declaring local
> variables as close as possible to their first use:
> > https://barrgroup.com/72-initialization#footnote12

The footnote refers to [Uwano], which can be found here:
[Uwano]: https://www.cs.kent.edu/~jmaletic/Prog-Comp/Papers/Uwano06.pdf

> 
> Hmm... seems  they don't provide any data to back up their statements.
> Specially that one sounds weird for me:
> " Too many programmers assume the C run-time will watch out for them, e.g., by
> zeroing the value of uninitialized variables on system startup."
> Why on earth people would assume that?

Not all programmers remember all the rules all the time. Especially junior developers.

> And what exactly means 'too many? 1%? 10%? 90%?

I guess that "too many" means that it is a statistically significant cause of bugs.

PS:
I like your way of reasoning.
I guess the Barr Group is trying to keep it short in their handbook, omitting the details from the underlying research.
It's a shame Jack Ganssle stopped giving his "How to Develop Better Firmware Faster" seminar (https://www.ganssle.com/classes.htm). All his "rule-of-thumb" guidelines are backed with hard data from references and experiments!

> 
> >
> > Old people (like myself) need to unlearn their bad old habits (originating
> from limitations in old C standards), and embrace modern
> > methods to reduce the risk of introducing bugs.
> 
> Allowing to define variables in the middle of the code by itself wouldn't
> prevent of use of un-initialized variables.
> From other side - compilers are quite good these days to catch such bugs.
> So I don't think it is a completing argument..

Please note that I am talking about "effectively uninitialized" variables,
meaning variables that have been initialized with dummy values like NULL, 0 or -1,
only to make the "use of uninitialized variable" compiler warnings go away.

Initializing variables with dummy values effectively disables the compiler's ability to catch bugs where a variable is being used before it has been assigned a (correct) value, because the compiler cannot know that the variable has been initialized with a dummy value.

The advantages of declaring the variable where it is used the first time are:
- The developer is much likelier to assign it the correct value to begin with.
- The reviewer is much likelier to spot if it is initialized with an incorrect value.


  reply	other threads:[~2024-06-20  9:02 UTC|newest]

Thread overview: 11+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-06-10 15:10 Morten Brørup
2024-06-10 16:11 ` Tyler Retzlaff
2024-06-10 16:31 ` Konstantin Ananyev
2024-06-20  0:38   ` Thomas Monjalon
2024-06-20  7:53     ` Morten Brørup
2024-06-20  8:09       ` Konstantin Ananyev
2024-06-20  9:02         ` Morten Brørup [this message]
2024-06-20 14:45           ` Stephen Hemminger
2024-06-11 15:10 ` Ferruh Yigit
2024-06-11 15:50   ` Stephen Hemminger
2024-06-17 14:38 ` Bruce Richardson

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=98CBD80474FA8B44BF855DF32C47DC35E9F540@smartserver.smartshare.dk \
    --to=mb@smartsharesystems.com \
    --cc=bruce.richardson@intel.com \
    --cc=david.marchand@redhat.com \
    --cc=dev@dpdk.org \
    --cc=ferruh.yigit@amd.com \
    --cc=konstantin.ananyev@huawei.com \
    --cc=stephen@networkplumber.org \
    --cc=thomas@monjalon.net \
    /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).