[OpenJDK 2D-Dev] <AWT Dev> PING: RFR: 8249215: JFrame::setVisible crashed with -Dfile.encoding=UTF-8

[Yasumasa Suenaga]

On 2020/07/30 12:20, Sergey Bylokhov wrote:
> On 29.07.2020 19:37, Yasumasa Suenaga wrote:
>> On 2020/07/30 11:03, Sergey Bylokhov wrote:
>>> On 28.07.2020 18:52, Yasumasa Suenaga wrote:
>>>> Hi Phil, Sergey,
>>>>
>>>> Sorry, I missed. null is WComponentPeer::pData, not defaultFont.
>>>
>>> Could you please clarify in what exact place access to the null pData in WComponentPeer caused a crash?
>>
>> src/java.desktop/windows/native/libawt/windows/awt_Component.cpp
>>
>> 6032         AwtFont *awtFont = (AwtFont *)env->GetLongField(font, AwtFont::pDataID);
>> 6033         if (awtFont == NULL) {
>> 6034             /*arguments of AwtFont::Create are changed for multifont component */
>> 6035             awtFont = AwtFont::Create(env, font);
>> 6036         }
>> 6037         env->SetLongField(font, AwtFont::pDataID, (jlong)awtFont);
>> 6038
>> 6039         c->SetFont(awtFont); // <--
>>
>> awtFont in above is NULL.
> 
> So the reason is that AwtFont::Create returns NULL, and if you take a look to that method
> there are a few cases when we could return NULL or return 0. Should not all that "returns"
> be removed? or all results of its usage wrapped by the "if(awtFont!=null)".
> Or it is fine as-is?

I think it is fine as-is with my proposal (webrev.03).

AwtFont::Create() would call GetComponentForName(), then it returns component font name from WFontConfiguration::getTextComponentForName. It is the method what I fixed in this change!


Thanks,

Yasumasa