From: Matthieu Moy <matthieu.moy@grenoble-inp.fr>
Date: Fri, 18 Jun 2010 06:30:33 GMT
Subject: Re: What's cooking in git.git (Jun 2010, #02; Wed, 16)
Message-ID: <vpq8w6cc0om.fsf@bauges.imag.fr>
In-Reply-To: <7vmxut1b8v.fsf@alter.siamese.dyndns.org>

Junio C Hamano <gitster@pobox.com> writes:

> Clément Poulain <clement.poulain@ensimag.imag.fr> writes:
>
>> A fix was proposed here:
>> http://mid.gmane.org/1276610328-28532-1-git-send-email-axel.bonnet@ensimag.imag.fr
>
> This one changes textconv_object() to take a pointer to ulong*; a caller
> is changed to cast a long* down to that pointer (in
> fill_origin_blob);

Isn't sign-casting safe? If not, then the existing call to
read_sha1_file should be fixed too:

-file->ptr = read_sha1_file(o->blob_sha1, &type,
-   (unsigned long *)(&(file->size)));
[...]
+file->ptr = read_sha1_file(o->blob_sha1, &type,
+   (unsigned long *)(&(file->size)));

(it's been there for a while)

>     The only portable way in C to convert between types is by assignment. So
>     you have to do:
>
>       unsigned long foo;
>       textconv_object(read_from, null_sha1, &buf.buf, &foo);
>       buf.len = foo;

This is what the patch does a little later:

+unsigned long buf_len;
[...]
+if (DIFF_OPT_TST(opt, ALLOW_TEXTCONV) &&
+    textconv_object(read_from, null_sha1, &buf.buf, &buf_len))
+buf.len = buf_len;

-- 
Matthieu Moy
http://www-verimag.imag.fr/~moy/