Bug 24912 – [Linux] Webkit incompatible with Java plugins

• Component: WebKitGTK
• Status: RESOLVED FIXED

[Samuel Creshal]

None of the Java plugins available for linux (x64) seems to be compatible with WebkitGtk, you get error messages like "libjavaplugin.so: Undefined symbol: NP_GetValue" when trying to load them.

[javier]

I have midori and libwebkit installed under arch-linux, and right now it looks like the sun java plugin gets recognized.  Actually when looking at "Tools -> Netscape plugins" I can see the sun java one information...

The problem is that even though that, when going to java.com, and trying to detect the java support, it doesn't get detected, and when going to other web pages requiring java support, they fail (such as my bank one)...

So I believe there's still problems, but not the same...  Versions on archlinux:

midori 0.1.7-1
libwebkit 1.1.8-2

[javier]

Hmm, well my previous comment stands for x86_64.  For i686, midori browser just crashes...

[cihat ertem]

I have the same problem with midori and webkit(midori 1.7.0-0, webkit 1.1.10-1). I am using ubuntu 9.04 64bit and both sun-java6-plugin and icedtea6-plugin seem useless when testing java versin in http://www.java.com/en/download/help/testvm.xml and http://www.javatester.org/version.html. I had opened a ticket at midori buglist/forum and they answered to me the problem was about webkitgtk. well, I see there is the problem for 4 months and it is still as "UNCONFIRMED" and no answer.

thanks and kind regards

[javier]

Well, with libwebkit 1.1.10-1 now the i686 behavior matches the x86_64 one.  Too bad still no java support though...

thx,

[Maciej Piechotka]

With webkit 1.1.19 and icedtea6-1.7:
Information that java failed to start:
** (epiphany:26620): DEBUG: NP_Initialize
** (epiphany:26620): DEBUG: NP_Initialize succeeded
** (epiphany:26620): DEBUG: NP_Initialize
** (epiphany:26620): DEBUG: NP_Initialize succeeded
[XPCOM] Failed to get the service manager
** (epiphany:26620): DEBUG: NP_Initialize
** (epiphany:26620): DEBUG: NP_Initialize succeeded
/var/tmp/paludis/dev-java-icedtea6-1.7/work/icedtea6-1.7/plugin/icedteanp/IcedTeaNPPlugin.cc:1427: thread 0x262cde0: Error: Failed to spawn applet viewer: Failed to execute child process "/usr/share/java-config-2/nsplugin/../../bin/java" (No such file or directory)
** (epiphany:26620): DEBUG: NP_Initialize
** (epiphany:26620): DEBUG: NP_Initialize succeeded
[XPCOM] Failed to get the service manager
** (epiphany:26620): DEBUG: NP_Initialize
** (epiphany:26620): DEBUG: NP_Initialize succeeded
** (epiphany:26620): DEBUG: NP_Initialize
** (epiphany:26620): DEBUG: NP_Initialize succeeded
[XPCOM] Failed to get the service manager
/var/tmp/paludis/dev-java-icedtea6-1.7/work/icedtea6-1.7/plugin/icedteanp/IcedTeaNPPlugin.cc:1427: thread 0x262cde0: Error: Failed to spawn applet viewer: Failed to execute child process "/usr/share/java-config-2/nsplugin/../../bin/java" (No such file or directory)

It works ok with firefox.

With webkit 1.1.19 and icedtea 1.6:
Just:
alt="Your browser understands the <APPLET> tag but isn't running the applet, for some reason."

[Xan Lopez]

Created attachment 48873 [details]
Add WebKitWebSetting to enable/disable Java plugin support.

[Xan Lopez]

Created attachment 48883 [details]
resolvesymlinks.diff

Resolve symlinks when loading plugins as modules.

[Xan Lopez]

Created attachment 48886 [details]
createjavaappletwidget.diff

This implements the createJavaAppletWidget, which turns out to be relevant here. Who could have guessed.

[Xan Lopez]

Created attachment 48892 [details]
npversion.diff

Make the supported NP version backend dependent to avoid insta-crash in the Java plugin.

[WebKit Review Bot]

Attachment 48892 [details] did not pass style-queue:

Failed to run "WebKitTools/Scripts/check-webkit-style" exit_code: 1
WebCore/plugins/gtk/PluginPackageGtk.cpp:149:  Place brace on its own line for function definitions.  [whitespace/braces] [4]
WebCore/plugins/win/PluginPackageWin.cpp:338:  Place brace on its own line for function definitions.  [whitespace/braces] [4]
WebCore/plugins/symbian/PluginPackageSymbian.cpp:177:  Place brace on its own line for function definitions.  [whitespace/braces] [4]
WebCore/plugins/mac/PluginPackageMac.cpp:305:  Place brace on its own line for function definitions.  [whitespace/braces] [4]
WebCore/plugins/wx/PluginDataWx.cpp:78:  Place brace on its own line for function definitions.  [whitespace/braces] [4]
WebCore/plugins/qt/PluginPackageQt.cpp:146:  Place brace on its own line for function definitions.  [whitespace/braces] [4]
Total errors found: 6


If any of these errors are false positives, please file a bug against check-webkit-style.

[Xan Lopez]

Created attachment 48897 [details]
npversion.diff

Should fix the style warnings.

[Gustavo Noronha (kov)]

Comment on attachment 48883 [details]
resolvesymlinks.diff

I should get credit for this part =P lol

[Gustavo Noronha (kov)]

Comment on attachment 48897 [details]
npversion.diff

I feel we're invading NPAPI's territory by using the NP prefix here. What do you think of supportedNPVersion?

[Xan Lopez]

Comment on attachment 48873 [details]
Add WebKitWebSetting to enable/disable Java plugin support.

Landed as r54886.

[Xan Lopez]

Comment on attachment 48883 [details]
resolvesymlinks.diff

Landed as r54887 giving credit to Gustavo :)

[Xan Lopez]

Comment on attachment 48886 [details]
createjavaappletwidget.diff

Landed as r54888.

[Xan Lopez]

CCing Maciej in case he wants to have a look at the versioning patch.

[Eric Seidel (no email)]

I'm confused, given the number of patches on this bug, if there is still action needed her or not?

[Eric Seidel (no email)]

Comment on attachment 48897 [details]
npversion.diff

Looks OK.

[Xan Lopez]

Comment on attachment 48897 [details]
npversion.diff

Landed in r54940.

[Xan Lopez]

All patches landed, closing bug.

[Alexey Proskuryakov]

+        uint16 NPVersion() const;

Why isn't this static?

> I feel we're invading NPAPI's territory by using the NP prefix here. What do
> you think of supportedNPVersion?

The NPVersion() function name also violates WebKit style.

[Xan Lopez]

(In reply to comment #22)
> +        uint16 NPVersion() const;
> 
> Why isn't this static?

Yeah, I actually thought of this, not sure why I didn't do it. I agree it makes sense.

> 
> > I feel we're invading NPAPI's territory by using the NP prefix here. What do
> > you think of supportedNPVersion?
> 
> The NPVersion() function name also violates WebKit style.

So what about 'supportedVersion'? Or 'supportedNPVersion' as Gustavo says?

[javier]

Hi, I'm using right now:

local/libwebkit 1.2.0-1
local/midori 0.2.4-1
local/jre 6u20-1

I think webkit 1.2.0 already grabbed the fixes on this bug, since before when trying to detect the java plugin on java.com with midori (webkit-gtk) was not possible.  Now it is, and the right version of the plugin is found by java.com.

While som things are working, still some are not.

The bank I have my accounts under uses something they call "virtual keyboard", it's a java applet which allows entering the letters of the password through the keyboard, but the numbers need to be button pressed by the mouse.  The applet shows up, and one can enter the password, but when pressing enter, it never gets to grant access to the bank account, although there's no error reported.  With firefox and the same version of the plugin, I have no problems at all.

The other issue I'm phasing is that under a corporate firewall, with firefox the java plugin is detected with no problem.  However with midori (webkit-gtk) the plugin is never detected, as the prior to these fixes behavior.  The java support should also be granted under a firewall.

I'm taking advantage the bug hasn't been closed, even though it says it's resolved/fixed.

Could any one look into these issues?  I know the bank account one is hard to reproduce since most people won't have a bank account under "www.bncr.fi.cr", but hey, I'm hoping some other similar applets might also be failing...

Also, is it recommended to open new bugs, or move this one to open, or perhaps just hope the issues are to be addressed with this same one?