New coverity run Feb 7, 14 new warnings

[Dan Kegel]

Congrats to everyone who's been cleaning up Coverity warnings,
especially Marcus (http://marcusmeissner.livejournal.com/24466.html).

Here are the yet-uninspected warnings that are in the Feb 7 run
but not in the Jan 30 run:

884     REVERSE_INULL   compile_file_regex      dlls/dbghelp/symbol.c
883     RESOURCE_LEAK   forward                 tools/winegcc/winegcc.c
882     RESOURCE_LEAK   build                   tools/winegcc/winegcc.c
881     RESOURCE_LEAK   build                   tools/winegcc/winegcc.c
880     RESOURCE_LEAK   build                   tools/winegcc/winegcc.c
879     RESOURCE_LEAK   compile                 tools/winegcc/winegcc.c
878     RESOURCE_LEAK   compile                 tools/winegcc/winegcc.c
877     RESOURCE_LEAK   compile                 tools/winegcc/winegcc.c
876     OVERRUN_STATIC  DOSVM_outport           dlls/winedos/ioports.c
875     NULL_RETURNS    __wine_make_gdi_object_system   dlls/gdi32/gdiobj.c
874     FORWARD_NULL    merge_diff_row          dlls/msi/database.c
873     FORWARD_NULL    HTMLStyle_put_fontWeight        dlls/mshtml/htmlstyle.c
872     FORWARD_NULL    NTDLL__vsnprintf        dlls/ntdll/printf.c
871     DEADCODE        reg_delete              programs/reg/reg.c

If anyone wants access to the Coverity data, let me or one of the other
coverity users know.
- Dan

[Alistair Leslie-Hughes]

Hi Dan,
> 873     FORWARD_NULL    HTMLStyle_put_fontWeight 
> dlls/mshtml/htmlstyle.c

NULL is a valid value to be passed though to gecko.

Best Regards
 Alistair Leslie-Hughes

[Dan Kegel]

Alistair wrote:
> > 873     FORWARD_NULL    HTMLStyle_put_fontWeight  dlls/mshtml/htmlstyle.c
>
> NULL is a valid value to be passed though to gecko.

Here's more of the error:

dlls/mshtml/htmlstyle.c

Event var_deref_model: Variable "v" tracked as NULL was passed to a
function that dereferences it. [model]
Also see events: [var_compare_op]

646  	        return set_nsstyle_attr(This->nsstyle, STYLEID_FONT_WEIGHT, v, 0);
647  	    }
648

htmlstyle.c:

216  	HRESULT set_nsstyle_attr(nsIDOMCSSStyleDeclaration *nsstyle,
styleid_t sid, LPCWSTR value, DWORD flags)
217  	{
218  	    nsAString str_name, str_value, str_empty;
219  	    LPWSTR val = NULL;
220  	    nsresult nsres;
221  	
222  	    static const PRUnichar wszEmpty[] = {0};
223

At conditional (1): "flags & 1 != 0" taking false path

224  	    if(flags & ATTR_FIX_PX)

Event deref_parm_in_call: Dereferenced parameter "value" in the
function "fix_px_value" [model]
Also see events: [deref_parm_in_call]

225  	        val = fix_px_value(value);

At conditional (2): "flags & 2 != 0" taking true path

226  	    if(flags & ATTR_FIX_URL)

Event deref_parm_in_call: Dereferenced parameter "value" in the
function "fix_url_value" [model]
Also see events: [deref_parm_in_call]

227  	        val = fix_url_value(value);

So it's not passing it to gecko, it's passing it to another mshtml function
that maybe passes it to fix*value() which dereferences it.

Is it a bug?  I don't know those flags well enough.
- Dan

[Jacek Caban]

Dan Kegel wrote:
> Is it a bug? I don't know those flags well enough.

It's not a bug, a constant 0 value is passed as flags to 
set_nsstyle_attr. I've marked it as FALSE in coverity.

Jacek