All of lore.kernel.org
 help / color / mirror / Atom feed
From: "Krzysztof Kowalczyk" <kkowalczyk@gmail.com>
To: "Jeff King" <peff@peff.net>
Cc: git@vger.kernel.org, gitster@pobox.com
Subject: Re: [PATCH] Optimize common pattern of alloc_ref from string
Date: Sat, 10 May 2008 17:30:56 -0700	[thread overview]
Message-ID: <7ce338ad0805101730n5b964a0em39d9fdcd9fc45f00@mail.gmail.com> (raw)
In-Reply-To: <20080510233918.GA315@sigill.intra.peff.net>

On Sat, May 10, 2008 at 4:39 PM, Jeff King <peff@peff.net> wrote:
> On Sat, May 10, 2008 at 04:26:58PM -0700, kkowalczyk@gmail.com wrote:
>
>> As a byproduct, fixes one place where string wasn't properly terminated.
>
> Great. Does this fix a user-visible bug? It would be nice to mention in
> the commit log _which_ place (though after reading the patch carefully,
> it looks like the one interpret_target) so that people looking at the
> commit later can understand exactly what was fixed.

It was a subtle memory corruption that wouldn't cause problems in
99.99% cases, but valgrind would probably catch it. And yes, it's the
interp_target().

>> -     ref = alloc_ref(strlen(refname) + 1);
>> -     strcpy(ref->name, refname);
>> +     ref = alloc_ref_from_str(refname);
>
> So this turns a 2-line construct into a 1-line construct...

And avoids future prossible mistakes with not terminating the string,
like the one just commited.

>> +struct ref *alloc_ref_from_str(const char* str)
>> +{
>> +     struct ref *ret;
>> +     unsigned len = strlen(str) + 1;
>> +     char *tmp = xmalloc(sizeof(struct ref) + len);
>> +     ret = (struct ref*)tmp;
>> +     memset(tmp, 0, sizeof(struct ref));
>> +     tmp += sizeof(struct ref);
>> +     memcpy(tmp, str, len);
>> +     return ret;
>> +}
>
> But why do we need an 8-line function to do it?
>
> The only difference I can see over
>
>  struct ref *alloc_ref_from_str(const char *str)
>  {
>    unsigned len = strlen(str) + 1;
>    struct ref *ret = alloc_ref(len);
>    memcpy(ret->name, str, len);
>    return ret;
>  }
>
> is that we avoid memsetting the name portion of the struct to 0 before
> copying to it. It seems like an unproven micro-optimization that makes
> it a bit harder to read.

You're absolutely right - it's a micro-optimization and your version
might be preferred for clarity. This is the first time I submit a
patch to git so I don't have a good feel for what kind of treadoffs
people find acceptable.

I should also mention that
static struct ref *try_explicit_object_name(const char *name)
{
	unsigned char sha1[20];
	struct ref *ref;

	if (!*name) {
		ref = alloc_ref(20);
		strcpy(ref->name, "(delete)");
		hashclr(ref->new_sha1);
		return ref;
	}
...

could also be replaced with alloc_ref_str() - I just wasn't 100% sure
if overallocating 10 bytes (20 - strlen("(delete)")) was just sloppy
code or does other code relies on that (which is unlikely and if true
then it wouldn't be good).

Regards,

-- kjk

  reply	other threads:[~2008-05-11  0:31 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2008-05-10 23:26 [PATCH] Optimize common pattern of alloc_ref from string kkowalczyk
2008-05-10 23:39 ` Jeff King
2008-05-11  0:30   ` Krzysztof Kowalczyk [this message]
2008-05-11  8:07     ` Jeff King
2008-05-11 18:39       ` Junio C Hamano
2008-05-11 20:15         ` Krzysztof Kowalczyk
2008-05-11 22:02           ` Junio C Hamano
2008-05-11 15:32 ` Johannes Schindelin

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=7ce338ad0805101730n5b964a0em39d9fdcd9fc45f00@mail.gmail.com \
    --to=kkowalczyk@gmail.com \
    --cc=git@vger.kernel.org \
    --cc=gitster@pobox.com \
    --cc=peff@peff.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.