Re: [ros-dev] [ros-diffs] [hbelusca] 73895: [NTGDI]: Few fixes & hacks for NtGdiGetGlyphIndicesW, see CORE-12825: - (Fix) Check for integer count overflow (per Thomas suggestion), - (Hack#1) Signal the particular calling ca...

[Thomas Faber]

On 2017-02-24 02:52, hbelusca(a)svn.reactos.org wrote:
>   @@ -4419,12 +4422,31 @@
>       LPCWSTR UnSafepwc = pwc;
>       LPWORD UnSafepgi = pgi;
>  -    if ((!UnSafepwc) && (!UnSafepgi)) return cwc;
>  -
>  -    if ((UnSafepwc == NULL) || (UnSafepgi == NULL))
>  +    /* Check for integer overflow */
>  +    if (cwc & 0x80000000) // (INT_MAX + 1) == INT_MIN
>  +        return GDI_ERROR;
>  +
>  +    if (!UnSafepwc && !UnSafepgi)
>  +        return cwc;
>  +
>  +    if (!UnSafepwc || !UnSafepgi)
>       {
>           DPRINT1("UnSafepwc == %p, UnSafepgi = %p\n", UnSafepwc, UnSafepgi);
>  -        return -1;
>  +        return GDI_ERROR;
>  +    }
>  +
>  +    // TODO: Special undocumented case!
>  +    if (!pwc && !pgi && (cwc == 0)) 
Why the sudden switch away from using the Unsafe version?
>   +    {
>  +        DPRINT1("ERR: NtGdiGetGlyphIndicesW with (!pwc && !pgi &&
> (cwc == 0)) is UNIMPLEMENTED!\n");
>  +        return 0;
>  +    }
>  +
>  +    // FIXME: This is a hack!! (triggered by e.g. Word 2010). See CORE-12825
>  +    if (cwc == 0)
>  +    {
>  +        DPRINT1("ERR: NtGdiGetGlyphIndicesW with (cwc == 0) is
> UNIMPLEMENTED!\n");
>  +        return GDI_ERROR;
>       }
>       dc = DC_LockDc(hdc);
>

[Neo Love]

If 'cwc' is a pointer or anything that would ever become 64 bits in x64
this type of coding needs rewriting using symbolic sizes.
(Such as ULONG_PTR(-1) and similar constructs..)
On 2017-02-24 20.42, Thomas Faber wrote:
>   +    /* Check for integer overflow */
>  +    if (cwc & 0x80000000) // (INT_MAX + 1) == INT_MIN
>  
L.

[Hermès BÉLUSCA-MAÏTO]

Hi,
I completely agree, and that's why I put in comment that INT_MIN stuff.
Hermès
-----Message d'origine-----
De : Ros-dev [mailto:[e-mail]] De la part de Neo Love
Envoyé : vendredi 3 mars 2017 01:26
À : ReactOS Development List
Objet : Re: [ros-dev] [ros-diffs] [hbelusca] 73895: [NTGDI]: Few fixes & hacks for
NtGdiGetGlyphIndicesW, see CORE-12825: - (Fix) Check for integer count overflow (per
Thomas suggestion), - (Hack#1) Signal the particular calling ca...
If 'cwc' is a pointer or anything that would ever become 64 bits in x64 this type
of coding needs rewriting using symbolic sizes.
(Such as ULONG_PTR(-1) and similar constructs..)
On 2017-02-24 20.42, Thomas Faber wrote:
>   +    /* Check for integer overflow */
>  +    if (cwc & 0x80000000) // (INT_MAX + 1) == INT_MIN 
L.
_______________________________________________
Ros-dev mailing list
Ros-dev(a)reactos.org
http://www.reactos.org/mailman/listinfo/ros-dev