<AWT Dev> Review Request: JDK-8179014 JFileChooser with Windows look and feel crashes on win 10 + jre-8 131

Sergey Bylokhov sergey.bylokhov at oracle.com
Sat May 6 01:35:54 UTC 2017

The fix looks fine. 
But I have a question about test, should it be run on win=10.0 only? I guess in this case it will be skipped on some new/other windows which will be released someday? 

----- philip.race at oracle.com wrote: 
OK +1 to this fix. Please follow RDP2 procedures to request this fix in 9 .. 
> -phil. 
> On 05/04/2017 02:55 AM, Prem Balakrishnan wrote: 


Thanks Kevin and Phil for the review. 

I shouldn't have missed the 'break' statement. Mistake on my part :( 

The crash fix null check is in thrid switch case. 

As a good practice of making the method robust, I added the null check in first switch case as well. 

The null checks have been added in method: 

static jstring jstringFromSTRRET(JNIEnv* env, LPITEMIDLIST pidl, STRRET* pStrret) 

This method is invoked from below methods in ShellFolder2.cpp 




The crash was due to function call getDisplayNameOf() --- which gets fixed if we add null check only in third swich case. 

The null check in first switch case is more of making the code robust against possible crash. 

I examined calls to these native methods getDisplayNameOf(), doGetColumnValue() & createColumnInfo() - 

Only getDisplayNameOf() return has null checks on java side. There are no checks on java side for string being null for other two methods. 

These two methods are invoked from FilePane.java class, which is used only in XXXXFileChooserUI classes. 

I haven't seen any side effect in my jtreg test runs of JFileChooser with this fix. 

The null check for first switch statement never existed and there were no crashes reported. What I have done is more of preventive check. 

Request you to review. 





From: Phil Race 
> Sent: Saturday, April 29, 2017 11:00 PM 
> To: Kevin Rushforth 
> Cc: Prem Balakrishnan; Alexander Scherbatiy; Sergey Bylokhov; awt-dev at openjdk.java.net 
> Subject: Re: <AWT Dev> Review Request: JDK-8179014 JFileChooser with Windows look and feel crashes on win 10 + jre-8 131 

I agree. This switch statement clearly was written in the expectation that each case executed return. 

And as I said offline pls confirm that you are confident that all return paths actually handle that null which may not have been tested in practice before now. There seemed to be several code paths. 


> On Apr 29, 2017, at 5:33 AM, Kevin Rushforth < kevin.rushforth at oracle.com > wrote: 

I'm not a "R"eviewer for AWT, but this part of the change looks wrong to me: case STRRET_CSTR : +            if (pStrret->cStr != NULL) { return JNU_NewStringPlatform(env, reinterpret_cast<const char*>(pStrret->cStr)); +            } case STRRET_OFFSET : 

> Did you really intended that a NULL pStrret->cStr pointer should fall through to the next case (STRRET_OFFSET)? If so, then a comment is in order because this seems odd. 
> -- Kevin 
> Prem Balakrishnan wrote: 

Hi , 

Please review fix for JDK 9, 

Bug: https://bugs.openjdk.java.net/browse/JDK-8179014 

Webrev: http://cr.openjdk.java.net/~pkbalakr/8179014/webrev.00/ 


JFileChooser with Windows LAF crashes on Windows 10. 

This issue is reproducible only on Windows 10 (v1703). 

Creating GodMode directory is the only way we are able to reproduce this issue. 


IShellFolder::GetDisplayNameOf() method is used to retrieve the display name of folder/subfolder using the GUID. 

This method returns NULL for GodMode directory [i.e., folder created with name: GodMode.{ED7BA470-8E54-465E-825C-99712043E01C}]. 

In previous version of windows (tested on v1511) this call returns name “GodMode”. 

Hence we get crash when processing the return data from the this method. 

Why only on Windows LAF? 

Unlike other LAFs in Windows LAF “useSystemExtensionHiding” is set to true. 

This check avoids the call to IShellFolder::GetDisplayNameOf() method in all other LAF’s. 

Hence crash is not reported on other LAFs. 


Added missing NULL checks in native method. 

This results in native method returning NULL in this case, which has already been handled on java side, hence no additional change is needed. 



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/awt-dev/attachments/20170505/8c1ead08/attachment.html>

More information about the awt-dev mailing list