threads / discuss / 24232

Purpose of "struct object_entry *oe = oe;"?

Subject: Purpose of "struct object_entry *oe = oe;"?

## tl;dr

4 messages between Jun 29, 2010 and Jul 2, 2010.

replies: 3people: 3as markdown or json

Bruce Stephens· Jun 29, 2010, 12:28 UTC · lore

Out of curiousity, where did this convention/idiom come from, and what's it for?

I presume it's to remove a warning ("oe might be used uninitialised") on a compiler (or something) that's clever enough to attempt such analysis but too stupid to notice that the supposed initialisation is using something uninitialised (or perhaps the compiler's deliberately recognising the convention?). Is that right, or does it actually do something more?

I'm mostly surprised that it surprises me. Is it used commonly in other projects? (It appears not to be mentioned in CodingGuidelines; should it be?)

(First instance in git that I can see is 67affd5173da059ca60aab7896985331acacd9b4, 2006.)

Joshua Juran· Jul 2, 2010, 01:43 UTC · re: Bruce Stephens · lore

Re: Purpose of "struct object_entry *oe = oe;"?

On Jun 29, 2010, at 5:28 AM, Bruce Stephens wrote:

Show 17 quoted lines

> Out of curiousity, where did this convention/idiom come from, and  
> what's
> it for?
>
> I presume it's to remove a warning ("oe might be used  
> uninitialised") on
> a compiler (or something) that's clever enough to attempt such  
> analysis
> but too stupid to notice that the supposed initialisation is using
> something uninitialised (or perhaps the compiler's deliberately
> recognising the convention?).  Is that right, or does it actually do
> something more?
>
> I'm mostly surprised that it surprises me.  Is it used commonly in  
> other
> projects?  (It appears not to be mentioned in CodingGuidelines; should
> it be?)

I was wondering this myself. My compiler complains in (and only in) the *presence* of this idiom, rather than its absence, so I undid it:

jj/warn/uninitialized branch on GitHub http://github.com/jjuran/git/tree/jj/warn/uninitialized

jj/warn/uninitialized commit on GitHub
http://github.com/jjuran/git/commit/ 
8affbf2d8e46fbd5f3d6898aa07ea8548432e7bc
Signed-Off-By:  Joshua Juran <jjuran@gmail.com> (or s/gmail/metamage/)

A compiler that warns of uninitialized usage *unless* a variable is initialized *with itself* is doubly broken.

Cheers, Josh

David Aguilar· Jul 2, 2010, 08:48 UTC · re: Bruce Stephens · lore

Re: Purpose of "struct object_entry *oe = oe;"?

On Tue, Jun 29, 2010 at 01:28:19PM +0100, Bruce Stephens wrote:

Show 16 quoted lines

> Out of curiousity, where did this convention/idiom come from, and what's
> it for?
> 
> I presume it's to remove a warning ("oe might be used uninitialised") on
> a compiler (or something) that's clever enough to attempt such analysis
> but too stupid to notice that the supposed initialisation is using
> something uninitialised (or perhaps the compiler's deliberately
> recognising the convention?).  Is that right, or does it actually do
> something more?
> 
> I'm mostly surprised that it surprises me.  Is it used commonly in other
> projects?  (It appears not to be mentioned in CodingGuidelines; should
> it be?)
> 
> (First instance in git that I can see is
> 67affd5173da059ca60aab7896985331acacd9b4, 2006.)

There are some good explanations in the list archives. http://thread.gmane.org/gmane.comp.version-control.git/133278/focus=133422 http://thread.gmane.org/gmane.comp.version-control.git/124676/focus=124803

-- 
		David

Bruce Stephens· Jul 2, 2010, 09:28 UTC · re: David Aguilar · lore

Re: Purpose of "struct object_entry *oe = oe;"?

David Aguilar <davvid@gmail.com> writes:

[...]

> There are some good explanations in the list archives.
> http://thread.gmane.org/gmane.comp.version-control.git/133278/focus=133422
> http://thread.gmane.org/gmane.comp.version-control.git/124676/focus=124803

Thanks.  That matches my guess.

Wouldn't it be useful to mention it in CodingGuidelines?

A danger is that the declaration (that yes, this really is initialised before use) becomes invalid during further changes, so it's best if everybody concerned understands the convention.

Obviously if it's a commonly understood and used convention then it's just as well not to have it in CodingGuidelines, but I'm not convinced that this one is, though (having noticed it) the intention isn't hard to grok.