48483 - Get the default plugin for Unix platforms

Status: VERIFIED FIXED

(Core Graveyard :: Plug-ins, defect, P1)

Description

[shrirang khanzode]

I am logging this bug to follow up the addition of default plugin for linux and 
mac. Windows already has this in the builds.

Comment 1

[shrirang khanzode]

*** Bug 48431 has been marked as a duplicate of this bug. ***

Comment 2

[J.J. Enser]

When you have figured this out, let me know the release/packaging implication if 
any. Thanks.

Comment 3

[leger]

Putting on nsbeta3 nominee radar.

Comment 4

[Marc Attinasi]

Approving for nsbeta3+ per prior discussion.

Comment 5

[shrirang khanzode]

*** Bug 46663 has been marked as a duplicate of this bug. ***

Comment 6

[av (gone)]

*** Bug 49381 has been marked as a duplicate of this bug. ***

Comment 7

[Viswanath Ramachandran]

I'm still seeing this bug in todays windows commercial build. looks like this is 
missing in windows too. 

Andrew Volkov wrote:
That means the installation lacks the default plugin which is supposed to take 
over in a case 
when user has not installed a plugin for dispalying given context. The default 
plugin should be 
part of distribution so ideally this message should never be seen. It is already 
included into 
Windows package and there is nsbeta3+ bug for not having it on Unix and Mac. 
Andrei Eric Krock wrote: Andrei? Viswanath Ramachandran wrote:   
what is this "Netscape Cant Find the Plugin Downloader Plugin" message that pops 
up when you read this email in Seamonkey? thanks, 
   Vishy 

Comment 8

[shrirang khanzode]

I did a typical install using today's build. Windows build packages the default 
plugin 'npnul32.dll'. I see the replacement puzzle icon on windows(due to 
default plugin)in today's build (2000082308m18)

Comment 9

[av (gone)]

I think what vishy is talking about is a different thing. It is about not seeing 
plugins in _mail_. Shrirang, could you please check it and file a bug if needed?

Comment 10

[Phil Peterson]

PDT agrees P1

Comment 11

[av (gone)]

*** Bug 45923 has been marked as a duplicate of this bug. ***

Comment 12

[Asa Dotzler [:asa]]

*** Bug 49121 has been marked as a duplicate of this bug. ***

Comment 13

[leger]

Putting [pdtp1] in whiteboard

Comment 14

[av (gone)]

*** Bug 52355 has been marked as a duplicate of this bug. ***

Comment 15

[drapeau]

I can help with this bug.  Re-assigning to me.

Comment 16

[av (gone)]

Adding myself to the cc list. Thank George for your help.

The bug is essentially about grabbing old code for the default plugin and put it 
in the Mozilla tree. Actions items as I see it:

1. Place the code under modules/plugin/default/[platform]
2. Compile and test it
3. Fix if it doesn't work. I think Unix version will require a trivial fix
4. Check it in
5. Modify appropriate make files to add it to the build process
6. Don't forget to modify packages so it gets to distribution (commercial 
tree) too

I can create an attachment with zipped code if needed.

Comment 17

[drapeau]

Working on this.  Thank you for the help, Andrei.

Comment 18

[Stephen Mak]

got the 4.x nullplugin code, created the makefile.in and integrated into
the netscape6 tree (solaris). However, it crashed and found that the code
use motif. Need to rewrite the part with Gtk. Estimate need 2 more days in
order to finish the coding & testing (solaris/linux). Also, as we don't have
testing platform for mac, will need help from guys who familiar on the mac 
development platform to test as well.

one more question for andrew, there are 3 npunix files in the 4.x nullplugin 
code u gave to George, which one should i use? default/common/npunix.c, 
default/common/npunix2.cpp or default/default/common/npunix.cpp ?

Stephen MAK 

Comment 19

[av (gone)]

cc'ing polmann. Eric, can you answer this question, please?

Comment 20

[Eric Pollmann]

Based on age and revisions of the files in the ns repository, my best guess is
that you should use default/common/npunix.c.  I think the .cpp files were an
attempt at a c++ rewrite a few years back which was dropped on the floor.

Comment 21

[Stephen Mak]

while I am working on the Gtk+ dialog box. something come up in my mind.
actually, should I do the dialog box in Gtk+? or may be Netscape want to use
XUL so that the look/feel will be consistent? Also, it may possible instead
of using native Gtk+, netscape6 use something else intenally which warp
around Gtk+ library (nsgtk?). And can somebody point out where can i find some
sample dialog box code in the mozilla tree so that the look and feel will not
look so different as well. thanks.

Stephen MAK

Comment 22

[Michael La Guardia]

would not hold pr3 for this

Comment 23

[Eric Pollmann]

Stephen, I can't promise that will be simple, but it doesn't look too bad. 
There is code in mozila/extensions/wallet/src/wallet.cpp that may be of use. 
Look for the string nsIPrompt.  There is similar code in
mozilla/extensions/psm-glue/src/nsPSMUICallbacks.cpp

Comment 24

[av (gone)]

You can also look in modules/plugins/nsPluginHostImpl.cpp on how nsIPrompt 
dialog can be used. There are some concerns though with improper parenting 
of modal boxes of nsIPrompt type.

Comment 25

[Stephen Mak]

Hi,
    I re-wrote the motif dialog box with Gtk+, however, as I didn't fully
know what is the correct look & feel, I just follow what Netscape 4.x do
on unreconize mine type to pop up a simple dialog box (I don't know whether
it will be enough). Well, I just make it be simple and at least it works. 
  I tested on both Solaris and linux which works fine. A tar file is attached
and can I have somebody to review it? The major change I made is in 
file 'nullplugin.c' and you can different all the files with the original files
which gave by Andrew. Thanks.

Stephen MAK

Comment 26

[Stephen Mak]

Attachment: The proposed fix for this bug on unix platform

Comment 27

[Stephen Mak]

The attachment I just attached is a tar.gz file. should be called something
like nullplugin.tar.gz

Comment 28

[Stephen Mak]

opus, one more thing, as I don't have any Mac development environment, in order
to fix this bug, some Mac people need to help out and fix this bug on 
Mac platform as well.

[e-mail]

Comment 29

[timeless]

helpwanted: mac developers?
simon could you find someone who could help?

Comment 30

[Stephen Mak]

just found that the plugin/default/unix/Makefile.in I attached last time
is corrupted. resend it.

Stephen MAK

Comment 31

[Stephen Mak]

Attachment: correct version of mozilla/modules/plugin/default/unix/Makefile.in

Comment 32

[Stephen Mak]

Attachment: the latest nullplugincod fix with correct jvm plugin URL and correct the multiple dialog box problem (filename: nullplugincode.tar.gz)

Comment 33

[Stephen Mak]

Attachment: the diff version of the suggested change of mozilla/modules/plugin/makefile.in

Comment 34

[Stephen Mak]

Hi, 
  I just submitted 2 latest attachments for the proposed fix. I
  tested with the mozilla tree at 9/22/00 8:00pm on unix as well.
  The difference of my latest proposed fix (for unix platform) from 
  my previous one including:

  (i) for mine type application/x-java-vm, the null plugin will
      foward the browser to home.netscape.com/plugins/jvm.html now.
      (which will be the same as windows nullplugin)

  (ii)I fixed a minor bug which if more than one unknown plugins
      in a page, it will pop up more than one dialog box.

  Also, I still need some people to review and approve my bug fix before
I can check in the fix. It will be greatly appreciated if somebody
can review/approve it and give me feedback so that I can check in the fix
for unix platform. (this bug is nsbeta3+, right?)

  By the way, when i try to fix the bug, I found some related issues:

  1. Ed Burn (who made the change for the nullplugin code for windows)
     ask me to redirect the java mine type to -
     http://home.netscape.com/plugins/jvm.html.
     I found that the page only contain jvm plugins for windows, need to
     add the unix/mac version on that page as well.

  2. The original 4.x code will redirect any unknown plugins to

     http://cgi.netscape.com/cgi/plug-in_finder.cgi?<minetype>

     I tested the above URL and able to find the corresponding plugin for
     the specify unreconized minetype under widows (either using netscape6
     or IE).   however, when I tried to go that page on solaris (either
     using netscape4.x or mozilla), it always return '0 matching plug-in' for
     any unreconized mintype. No matter the minetype has a corresponding
     plug-in or not (i tested out the plug-ins avaliable on both solaris
     /win platform)

     It's kind of strange but I guess this should be logged as another bug.
     (it's because netscape4.x on solaris have the same proble when going
     to the above URL)

  3. The Mac version of the nullplugin code still need to be implemented and 
tested

  thanks,

Stephen MAK

Comment 35

[av (gone)]

Thank you Stephen. To get a review please follow the current review rules on the 
top of the Tinderbox page. I think at the moment we don't have much choice but 
send the diffs to designated people.

Comment 36

[Jason Eager]

Adding appropriate keywords and trying to get this bug a double ++ because the
null plugin is NOT available from netscape's site, and this causes all kinds of
confusion on the Linux side.

Comment 37

[Eric Pollmann]

We'll need to add this (or something like it) so that the Makefile will be
autogenerated by configure:

allmakefiles.sh
Index: allmakefiles.sh
===================================================================
RCS file: /cvsroot/mozilla/allmakefiles.sh,v
retrieving revision 1.250
diff -u -r1.250 allmakefiles.sh
--- allmakefiles.sh     2000/09/22 05:09:36     1.250
+++ allmakefiles.sh     2000/09/26 00:13:41
@@ -341,6 +341,7 @@
 modules/plugin/src/Makefile
 modules/plugin/test/Makefile
 modules/plugin/SanePlugin/Makefile
+modules/plugin/default/unix/Makefile
 "
 
 MAKEFILES_netwerk="

Also, we'll need to add the .so to mozilla/xpinstall/packager/packages-unix but
I'm not sure where it would go in that file.

Comment 38

[Eric Pollmann]

Just a few comments on the fix:

1) Wow, it's great to see this is finally being done!  :) thankyou thankyou...

2) The fix says:

27      /* this function is used to clear the mime type cache list 
28         whenever the dialog box is closed. We need to clear the
29         list in order to have the dialog box pop up again when
30         the page is reload. (it's because there is no puzzle
31         icon in unix to let the people actively forward to netscape
32         page)  */

Actually, in the old null plugin code, there is an equivalent of the puzzle
piece - just a plain old button that takes up the space where the plugin
appears.  (member variable "button")  This button is created in makeWidget(). 
The reason for this button is because we figured at the time that it would be
less intrusive than popping up a dialog box every time the user pressed Reload
or returned to the page clicking "Back" or "Forward".

Initially the button says: "Click here to get the plugin"  After it is clicked,
the message changes to "Click here after installing the plugin".  If it is
clicked after installing the plugin, the plugin list is refreshed (load the URL:
REFRESH_PLUGIN_LIST == "javascript:navigator.plugins.refresh(true)")  This
causes the newly installed plugin to be loaded immediately in place of the
button.

The status line of the browser is updated with the text of the button as the
mouse enters the button, and replaced as the mouse leaves the button.

All of the above functionality is "neat" but not needed.  The new implementation
as it is sounds fine.

3) I am not a GTK expert.  (standard disclaimer)  I reviewed all of the changes
to nullplugin.c, nullplugin.h, and npunix.c.  Based on my past experience with
the null plugin, and what little tiny grasp I have on GTK, it looks good.

I also tried compiling it on my FreeBSD machine, and with the above change
(previous post), everything compiled.

I also tested the plugin by going to:

http://www.apple.com/quicktime

The dialog box popped up, and clicking okay, opened up a new window with the
download form in it (sadly, the quicktime plugin is only available for Mac and
Windows.)

Looks good to me, r=pollmann

To check in this fix, please read the requirements on:
http://tinderbox.mozilla.org/showbuilds.cgi?tree=SeaMonkey

Comment 39

[Stephen Mak]

pollman, thank you so much for the review. 
I will go ahead to get approval, by the mean time, I just want to make sure 
the following change on packager/packages-unix is fine as well (u suggested it, 
right?)Index: packages-unix
===================================================================
RCS file: /cvsroot/mozilla/xpinstall/packager/packages-unix,v
retrieving revision 1.113
diff -u -r1.113 packages-unix
--- packages-unix       2000/09/20 19:35:19     1.113
+++ packages-unix       2000/09/26 01:16:12
@@ -208,6 +208,7 @@
 bin/defaults/pref/*
 ; this is only here if you have plugins -blizzard
 ; bin/plugins/*
+bin/plugins/libnullplugin.so
 bin/res/entityTables/*
 bin/defaults/wallet/FieldSchema.tbl
 bin/defaults/wallet/SchemaConcat.tbl
=======

Stephen MAK

Comment 40

[Stephen Mak]

One more thing, I tried to find a suitable reviewer to do the approval by 
checking the tinderbox page. However, I don't see any reviewer on the list
in the plug-in area. So, can anybody suggest who should I mail to in order
to get the approval for the bug fix? thanks so much.

Stephen MAK

Comment 41

[Brendan Eich [:brendan]]

Nits: please add an MPL comment header to all files, using the boilerplate at 
http://www.mozilla.org/MPL/boilerplate-1.1/.  Also expand tabs to the right (4 
space modulus?) tabstops, and configure the Emacs modeline comment at the top 
(in the boilerplate license comment, the very first line) appropriately.

I realize it's old code, but might as well make it consistent.  From npshell.c 
around line 98:

	This = (PluginInstance*) instance->pdata;
	memset(This, 0, sizeof(PluginInstance));

	if (This != NULL)
	{
	        ...

The memset should be moved inside the if (This != NULL) then-block.  Better, but 
bigger change: invert the sense of the test and return NPERR_OUT_OF_MEMORY_ERROR 
early, and have the normal code (that checks argn[i] for well-known names and 
stashes the values in This members) not be so overindented -- and avoid the else 
after return non-sequitur at line 134.

Unchecked This deref at line 177, in NPP_SetWindow.  I'm not sure if this is 
safe, given that This is null-checked in NPP_Destroy.  Old code again, but maybe 
we can improve it at this late date?

The else return len; at lines 251 and 252 in NPP_Write is goofy.

Under-indented code at lines 69-72 in nullplugin.c.

At line 65, instead of:

        char* buf = 
"javascript:netscape.softupdate.Trigger.StartSoftwareUpdate(\"%s\")";
        url = NPN_MemAlloc(strlen(This->pluginsFileUrl) + 1 + strlen(buf) + 1);
        if (url != NULL)
 
why not:

        static char buf[] = 
"javascript:netscape.softupdate.Trigger.StartSoftwareUpdate(\"%s\")";

        // subtract 2 for the %s, but leave the \0 counted by sizeof.
        url = NPN_MemAlloc((sizeof(buf) - 2) + strlen(This->pluginsFileUrl));
        if (url != NULL)
        {
            /* Insert the file URL into the JavaScript command */
            sprintf(url, buf, This->pluginsFileUrl);
            NPN_GetURL(This->instance, url, TARGET);
            NPN_MemFree(url);
        }
 
The sprintf at line 190 seems vulnerable to Morris-worm-style buffer overrun 
attacks, if the MIME type string can be sent from an attacking machine and have 
arbitrary length and byte-sized characters in it -- use PR_snprintf to clamp 
against buffer size, or if the null plugin does not link with NSPR, perhaps you 
can use a libc snprintf on most (all?) modern Unixes.

Blizzard would be a good guy to look over the gtk stuff, but it sounds like it 
all works great, from pollmann's testimony.

Thanks for doing this, please forgive the nit-picking and fix as many as you can 
and update the bug -- I'll say a=[e-mail] after a second look.

/be

Comment 42

[Stephen Mak]

Attachment: latest proposed fix which integrage the change suggested by Brendan (filename: nullplugincode.jar.gz)

Comment 43

[Stephen Mak]

hi,
   thanks Brendan for the suggestions. I have integrated all his suggestions
and submitted as attachment. Besides the news files needed to fix this bug, 
serveral makefile also need to change as well, here are the diffs of all the 
changes required (I tested and it works as before):
=========
Index: modules/plugin/Makefile.in
===================================================================
RCS file: /cvsroot/mozilla/modules/plugin/Makefile.in,v
retrieving revision 1.9
diff -u -r1.9 Makefile.in
--- modules/plugin/Makefile.in  2000/07/17 13:06:50     1.9
+++ modules/plugin/Makefile.in  2000/09/26 21:37:16
@@ -26,7 +26,7 @@

 include $(DEPTH)/config/autoconf.mk

-DIRS           = public nglsrc
+DIRS           = public nglsrc default/unix

 ifdef ENABLE_TESTS

-------------------------------------------------------------------------------
Index: xpinstall/packager/packages-unix
===================================================================
RCS file: /cvsroot/mozilla/xpinstall/packager/packages-unix,v
retrieving revision 1.113
diff -u -r1.113 packages-unix
--- xpinstall/packager/packages-unix    2000/09/20 19:35:19     1.113
+++ xpinstall/packager/packages-unix    2000/09/26 21:40:53
@@ -208,6 +208,7 @@
 bin/defaults/pref/*
 ; this is only here if you have plugins -blizzard
 ; bin/plugins/*
+bin/plugins/libnullplugin.so
 bin/res/entityTables/*
 bin/defaults/wallet/FieldSchema.tbl
 bin/defaults/wallet/SchemaConcat.tbl
--------------------------------------------------------------------------------
Index: allmakefiles.sh
===================================================================
RCS file: /cvsroot/mozilla/allmakefiles.sh,v
retrieving revision 1.250
diff -u -r1.250 allmakefiles.sh
--- allmakefiles.sh     2000/09/22 05:09:36     1.250
+++ allmakefiles.sh     2000/09/26 21:41:54
@@ -341,6 +341,7 @@
 modules/plugin/src/Makefile
 modules/plugin/test/Makefile
 modules/plugin/SanePlugin/Makefile
+modules/plugin/default/Makefile

-------------------------------------------------------------

thanks,

Stephen MAK

Comment 44

[Stephen Mak]

opus, a mistake in last post, the allmakefiles.sh diff should be:

--------------------------Index: allmakefiles.sh
===================================================================
RCS file: /cvsroot/mozilla/allmakefiles.sh,v
retrieving revision 1.250
diff -u -r1.250 allmakefiles.sh
--- allmakefiles.sh     2000/09/22 05:09:36     1.250
+++ allmakefiles.sh     2000/09/26 21:47:54
@@ -341,6 +341,7 @@
 modules/plugin/src/Makefile
 modules/plugin/test/Makefile
 modules/plugin/SanePlugin/Makefile
+modules/plugin/default/unix/Makefile
 "

 MAKEFILES_netwerk="
------------------------------------------

thanks,

smak

Comment 45

[Brendan Eich [:brendan]]

Looks good -- two more nits:

1.  The c-basic-offset should probably be 4 in all the .c and .h files.  And
there should be no tabs (run expand -4 over each file, diff -w the result to
make sure nothing important changed, and then mv the expand output file back
over the original file).

2.  static char buf[] = ... should be static const char buf[] = ...;

Fix these, do the diffs, and test a bit, and you get an automatic
a=[e-mail].

Comment 46

[Stephen Mak]

Attachment: minor change per Brendan's request. will be the final version before check-in (filename: nullplugincode.tar.gz)

Comment 47

[Michael La Guardia]

marking nsbeta3- per my earlier comment.  nominating for rtm.

Comment 48

[Stephen Mak]

I already checked in the fix to the trunk (unix platform fix). 
Per Jim Roskind's request, can anybody assign this bug to Eric Pollmann?

Stephen MAK
---------

> Assuming you get a review and land this on the trunk... and it works
> well... rather than closing the bug, would you mind assigning it to
> Pollman and marking "checked into trunk".  Also, please add the 
keyword
> "rtm" so that Netscape will consider taking the patch into the 
Netscape
> 6 branch.

> Thanks,

> Jim
> --

> -- My views are mine, not Netscape's --
> Jim Roskind                                              fax: 
650.428.4058
> [e-mail]                                       voice: 
650.937.2546

Comment 49

[Eric Pollmann]

--> self to migrate changes to the branch

Comment 50

[Eric Pollmann]

We still need to find an owner for the Mac part of this bug.  Andre?

Comment 51

[Jason Eager]

We still need to get this checked into the branch, otherwise there will be an
outpouring of frustration from Linux/unix users.

Comment 52

[Michael Marxmeier]

Now that Linux has a null plugin is there a way to _disable_ it?
Since a lot of plugins are not (yet) available for Linux having it
pop up each time is pretty annoying.
The message about the missing null plugin had a nice option to have
it shut up and AFAIR it dit not ask several times in a row.

Comment 53

[André Dahlqvist]

One observation on the Default plugin in 2000092821 on Linux:

Going to http://www.apple.com/quicktime will open the Default Plugin window
correctly, but the parenthesis after "type" will be empty. Clicking OK will
however still take you to the page where you can download quicktime.

Steps to reproduce:

1) Go to http://www.apple.com/quicktime

2) Notice that the parenthesis after type is empty:
   "This page contains information  of a type() that can only...

100% reproducable.

Comment 54

[Eric Pollmann]

[e-mail], I'll pull the changes over to the branch as soon as nsbeta3
goes out.  I'm waiting until then because this bug was marked nsbeta3-.

[e-mail], I also see this.  Will look into it before pulling the
changes over.  Stephen, do you have any ideas off the top of your head?

Comment 55

[André Dahlqvist]

One more comment using 2000092821 on Linux:

1) Go to http://www.macromedia.com

2) The Netscape Default Plugin will find flash.

3) As the new window opens Mozilla will crash.

Reproducable 99% of the time. (I managed to get the window to open once.)

Comment 56

[Kevin McCluskey (gone)]

Marking rtm+. Adding crash keyword based on [e-mail]'s comment.

Comment 57

[Kevin McCluskey (gone)]

Putting rtm+ in status not summary.

Comment 58

[Kevin McCluskey (gone)]

Marking [rtm need info] Waiting for module owner review.

Comment 59

[Stephen Mak]

I tried the mozilla build on my solaris box. However, when
i load the www.macromedia.com page, it doesn't pop up any
dialog box for plugins. (anyway, the page crashes around 2 out of 3 times
but not related to open any new window) I tried to remove my
null plugin code. It also doesn't pop up any dialog box about nullplugin
loader missing. Also asked my co-worker to test on linux with a 2 weeks
old build and have the same result.

Is the macromedia page changed or just do i miss anything? looks like there
are no plugin in the front page of macromeida.com

and for the apple.com/quicktime page, the main reason of missing type is 
because the apple page <embed> tag doesn't provide the plugin type attribute.
So, this is the reason why it's missing. May be the code should be change
a little bit so that in case there is no type information just skip it.

Stephen MAK

Comment 60

[Colin Blake]

I just tried building this on OpenVMS and when it gets to link libnullplugin.so 
I have an unresolved symbol: NPP_URLNotify.

I see that npunix.c makes this call. Where is NPP_URLNotify supposed to come
from?

Comment 61

[Colin Blake]

Sorry, should have said. This is using M18 (trunk) code I downloaded 10/2/2000, 
and OpenVMS builds "like UNIX".

Comment 62

[Stephen Mak]

more information about the macromedia page.
My mozilla browser will just crash because of the isprint() bug (bug#54694).
looks like it's not gonna to be fixed soon from bugzilla. 
I also tried on a 3 weeks old linux build. there is no plugin expected and I 
don't see any nullplugin dialog box pop up (macromedia change their homepage?). 
need to verify and check. 
also, anybody can verify with other plugin page to see whether it will crash?
(i only tried apple.com/quicktime and java.sun.com so far)

Comment 63

[Kevin McCluskey (gone)]

Marking rtm+. Have patch, super-review, checked in to the trunk.

Comment 64

[Kevin McCluskey (gone)]

Added reviewed, super-reviewed to status

Comment 65

[Stephen Mak]

colin,
    I used lxr to search for NPP_URLNotify and found that it's in
/plugin/oji/MRJ/plugin/Source/BackwardAdapter.cpp

Comment 66

[Colin Blake]

Stephen,

That's for the OJI plugin. I'm talking about the null plugin.

Colin.

Comment 67

[Phil Peterson]

PDT marking [rtm++]

Comment 68

[Eric Pollmann]

Colin, I think I see what you're referring to - should NPP_URLNotify be defined
in modules/plugin/default/unix/npshell.c where all of the other NPP_<foo>
functions are defined?  It seems like many of the functions don't do anything
(as has always been the case for the Un*x default plugin) so perhaps it would be
safest just to put a stub in there - if that's indeed where it belongs.

NPP_URLNotify is defined for the windows default plugin in npshell.cpp

Comment 69

[Eric Pollmann]

Attachment: attempt at OpenVMS patch

Comment 70

[Eric Pollmann]

Colin, can you try out the above patch to see if it fixes the problem you're
seeing on OpenVMS?  Thanks!

Comment 71

[Eric Pollmann]

[e-mail], luckily, the default plugin does not have anything to
do with the loading of the flash plugin.  The selection of a plugin to display a
given mime type is done by the plugin glue code, and only if no plugins for a
given type can be found is the default plugin loaded.  In fact, the flash plugin
was working and displaying flash on Linux long before the default plugin was
added.  That is to say a crash loading a flash page is probably not related to
this bug.

I also tried testing www.macromedia.com and did not see anything which invoked
the flash plugin, or any crashes.

I don't see any other outstanding issues with this bug, so if the patch above
makes OpenVMS happy, I'll pull it over to the branch (after one final review).

Comment 72

[Colin Blake]

With the 10/6 patch, the null plugin now builds cleanly. Note that this is NOT
an OpenVMS-specific fix; NPP_URLNotify is missing for all platforms, its just
that OpenVMS is one of the few O/S's that complains about missing symbols at
LINK time rather than at RUN time.

Comment 73

[Eric Pollmann]

Checked the last patch into the trunk, and everything into the branch.  Marking
this bug fixed.

To verify, use the latest Unix branch build to go to a page that has a plugin
that you don't have installed (http://apple.com/quicktime).  You should see a
dialog box pop up that says that you need to download a plugin, and when you
click Okay, it should take you to the download search page.

Comment 74

[Mike Pinkerton (not reading bugmail)]

and for mac?

Comment 75

[av (gone)]

Reopening as per Mac issue. Adding ekrock.

Comment 76

[av (gone)]

Taking it for now.

Comment 77

[av (gone)]

Now really taking.

Comment 78

[Eric Pollmann]

Sorry about the mixup - I asked Heikki to turn off the branch checkin last night
when it turned out I had broken HPUX (build on everything else, :S)  Just now I
turned the checkin back on thanks to some help form Jim D., Wan-Teh, and Scott
M.  Of course, that broke AIX.  But I just checked in a fix for *that* bustage. 
*whew*

Comment 79

[av (gone)]

Shrirang, could you please verify that the default plugin is now available for 
Unix and say so in the bug report? According to Eric it should appear in today's 
daily build.

Comment 80

[shrirang khanzode]

Yes, the default plugin(libnullplugin.so) is in the plugins folder on today's 
linux branch build 2000101211. However, I see other problems for which I will 
log seperate bugs.

Comment 81

[Stephen Mak]

pollmann,
  I just update the mozilla tree. and it seems that somebody update the npapi.h
file. I can't build it on my solaris 2.8 machine. it's because int16/32 & 
uint16/32 is not defined. I checekd and looks like  the #define _UINT16/32 & 
#define _INT16/32 in npunix.c/npshell.c/nullplugin.c no longer needed. I don't
know whether we should comment it out. and don't know whether it will affect 
other platform (aix/hpunix) as well....

Comment 82

[Eric Pollmann]

That was me.  :)  I updated npapi.h, npunix.c, npshell.c, and nullplugin.c
yesterday and the day before.  The changes were to fix build bustage on HPUX and
AIX.  AFAIK, if you update all of these files to the latest revs, they should
build cleanly on Solaris (2.6 and 2.7 tinderbox machines build cleanly).

A more complete outline of the change is this:
 - On HPUX, the combination of #defining _U/INT16/32 and #including npapi.h
   would leave int16 and int32 undefined because they are defined in model.h
   (included from obsolete/protypes.h) *only if* _U/INT16/32 are *not* defined.
 - To fix this, I removed the definitions of _U/INT16/32 from the *.c files and
   added #define NO_NSPR_10_SUPPORT to npapi.h.  This causes obsolete/protypes.h
   to not be included on any platform, and the definitions for int16/32 from
   npapi.h to be used on all platforms.  (Thanks wtc and jdunn for help there)
 - After checking this patch in, AIX was busted because AIX *already* gets
   definitions for int16/32 from sys/inttypes.h which is included in
   sys/types.h.  I put an #ifndef AIX around the definitions in npapi.h to
   prevent the double definition.

If you are experiencing problems with solaris 2.8 and have all of the latest
versions of these files with no local modifications, please let me know what the
errors are and fixes if you have them.  Thanks!

Comment 83

[Jim Roskind]

It sounds like the bug was fixed for linux, but not fixed for mac (and that was
why it was re-opened??)
Qa needs bugs marked fixed and rtm-double-plus to go and verify.  Having this
bug in limbo is bad (since we don't have reviewed fixes for mac, and we do have
a plus-plus :-/ ).  IF I marked it need-info, I'd loose the qa verification, but
leaving it plus-plus (and active) gives carte-blanche to folks to land :-(.

I'm going to mark this bug fixed, and ask folks that can isolate the additional
issues (mac?  solaris?) to open separate bugs.  It would be nicer if bugzilla
could handle this better ... but it can't.

Comment 84

[av (gone)]

Changing summary to specify Unix only, and a new bug for Mac is filed -- 56761

Comment 85

[timeless]

Changing OS to linux.

Comment 86

[shrirang khanzode]

This is fixed in branch commercial build 2000101609 on linux. Adding VTRUNK to 
get verified on trunk.

Comment 87

[Asa Dotzler [:asa]]

Verified Fixed on trunk build.  http://apple.com/quicktime page prompted with
get the plugin dialog and accepting ook me to download the player page.
linux 101808 RedHat 6.2
Setting bug to Verified and removing vtrunk keyword

Comment 88

[sairuh (rarely reading bugmail)]

*** Bug 49381 has been marked as a duplicate of this bug. ***